Datasets:
Dataset Viewer
pmid
string | section
string | text
string | annotations
sequence |
|---|---|---|---|
37299942 | intro | 1. Introduction Children who experience handwriting learning disabilities may face significant negative effects on their academic success and daily life. Cognitive, visual perceptual and fine motor skills are essential for learning to write. Dysgraphia refers to the motor performance difficulties of handwriting rather than the skills of spelling or text composition. Between 5 and 34% of children never master legible, fluent and enduring handwriting despite appropriate learning over a period of 10 years. Early detection of dysgraphia is important to initiate a targeted intervention for the improvement of the children’s writing skills, which are a basis for their academic career. Criterion-referenced assessment of handwriting skills in occupational therapy practice traditionally includes the quantitative and qualitative analysis of letter formation, spacing, alignment, legibility and handwriting speed. Recently, several studies have investigated automatic dysgraphia detection using machine learning algorithms with a digital tablet. Tablet sensors enable collecting signals, such as the x and y coordinates, pressure and tilt of the pen, during writing. However, these studies deployed classical machine learning algorithms, such as Random Forest4 and AdaBoost7, which require manual feature extraction and selection. Asselborn et al. extracted 53 handwriting features, which are classified into four groups: static, kinematic, pressure and tilt features. Spectrum features seem to be the most important features: six out of eight features are related to frequency. The four most discriminative features are the Bandwidth of Tremor Frequencies and the Bandwidth of Speed Frequencies (kinematic features), the Mean Speed of Pressure Change (pressure feature) and the Space Between Words (static feature). Using these 53 features and the random forest classifier, an F1-score of 97.98% was achieved. Accuracy was not reported due to an imbalanced data set (56 children with dysgraphia and 242 typically developing children). Asselborn et al. used PCA to reduce the number of relevant features and to enable a more data-driven classification. Drotar and Dobes extracted 133 features using mainly the statistical measures leak min, mean, median and standard deviation of features such as velocity, acceleration, jerk, pressure, attitude, azimuth, segment/vertical/horizontal length and pen lifts. Merged from all tasks, they produced 1176 features in total. From these 1176 features, 150 features were selected using weighted k-nearest-neighbour feature selection. Finally, using these 150 features as input to the AdaBoost classifier, an accuracy of 79.5% was obtained. The data set was almost balanced: 57 children with dysgraphia and 63 normally developing children. Dimauro et al. used 13 pure text-based features, such as writing size, non-aligned left margin, skewed writing and insufficient space between words, and achieved an accuracy of 96%. Nevertheless, the data were imbalanced, with 12 out of 104 children showing dysgraphia. In contrast, Devillaine et al. used only graphical tablet sensor signals, i.e., x, y and z positions and pressure data, for feature extraction and no static features from text data. From the extracted features, 10 features were selected using linear SVM or extra trees. Different classical machine learning algorithms were tested for classification of a balanced data set consisting of 43 children with dysgraphia and 43 normally developing children. The best performance with 73.4% accuracy was achieved with the random forest algorithm. Devillaine et al. also provided a comparison of different classical machine learning algorithms for dysgraphia detection, with the highest F1-score of 97.98% achieved by Asselborn et al.. To summarize, all the cited papers used feature extraction and selection, which require high effort and expertise. Furthermore, the feature extraction is also subjective, i.e., based on expert experience, which might miss some important features. In this project, an LSTM deep learning algorithm was used, which automatically extracts features from raw sensor signals. To the best of our knowledge, only a few research groups have deployed deep learning for dysgraphia detection, such as Ghause et al., achieving an F1-score of 98.16%. This work applied a CNN with text images as input, i.e., no dynamic features were used. Zolna et al. achieved a significantly better performance in dysgraphia detection when using an RNN (more than 90%) in comparison to a CNN (25–39%). Unlike a CNN, an RNN takes into account the dynamic aspects of the writing. However, they used only the trajectory of the consecutive points as dynamic features and not pressures, velocity, accelerations and tilt as we did. They also used two-layer LSTM but with 100 neurons in each layer and a dropout layer with 50% drop probability. In contrast to previous works, which used tablets, we used a pen equipped with sensors (SensoGrip) to capture handwriting dynamics (pressure, speed, acceleration and tilt), which looks and feels more like a real pen and enables writing evaluation in more realistic scenarios. The SensoGrip system consists of the SensoGrip pen and an app designed for the Android OS. The SensoGrip pen features an integrated microcontroller, which is able to communicate with the Android device via Bluetooth BLE. It also contains the necessary power supply, electronics and sensors for measuring tip and finger pressure, as well as an IMU MEMS three-axis accelerometer and three-axis gyroscope. The microcontroller captures the pressure data as well as the data provided by the IMU and forwards them to an app on the Android device. The user is able to acquire feedback via built-in RGB LEDs or via the mobile app. We do not provide binary output (dysgraphia or no dysgraphia) but the scores of the German version of the Systematic Screening for Motor Handwriting difficulties (SEMS). SEMS provides values between 0 and 12 for print and a maximum of 14 points for manuscript writing. More points indicate more difficulties in writing, and therefore, these scores enable a finer granularity in handwriting evaluation. In this work, we investigated the fine grading of handwriting capabilities by predicting the SEMS score (between 0 and 12) with deep learning. We achieved a root-mean-square error (difference between the predicted SEMS score and the SEMS score estimated by expert therapists) of less than 1, with automatic instead of manual feature extraction and selection. Furthermore, we used the SensoGrip smart pen, a pen equipped with sensors to capture handwriting dynamics, instead of a tablet, enabling writing evaluation in more realistic scenarios. In the course of this work, our results are presented and discussed, followed by a description of the applied methods. | [
[
1595,
1605
],
[
2027,
2047
],
[
3054,
3064
],
[
3800,
3810
],
[
4521,
4531
],
[
4757,
4767
],
[
5303,
5323
],
[
5529,
5539
],
[
6310,
6320
],
[
6887,
6897
],
[
6904,
6914
]
]
|
38861157 | results | Results The anterodorsal thalamic nucleus shows early and consistent vulnerability to tau pathology We examined post-mortem brain samples from 26 cases, grouped by the Braak tau stages of disease progression (Table 1). We assessed early (Braak stages 0–II; n = 15), middle (Braak stages III–IV, n = 6), and late (Braak stages V–VI, n = 5) cases, comparing the rostral thalamus (n = 25 cases) to the cortex (hippocampal formation and cingulate areas) and to other subcortical areas (Fig. 1a–f, 2, 3, 4; Tables 1, 2; Tables S3, S4). We quantified ptau coverage and ptau-immunoreactive (ptau+) cells using the AT8 antibody, which recognizes phosphorylated serine 202 and threonine 205 residues of tau, in 50-µm-thick sections from perfusion-fixed human brains using a pixel classifier (Fig. 1g, h; Figure S1; Table 1; Methods). We confirmed these data with a ptau-intensity scoring system in a larger number of samples comprising the perfusion-fixed sections and 10-µm-thick FFPE sections and FFIF sections (Fig. 1i, j; Table S3; n = 26 cases). We observed ptau+ neurons and processes restricted to the ADn in controls (‘pre-Braak’ stage 0; n = 4/6 cases; coverage = 0.56%, ptau+ cells = 1.29 mm−2, intensity score = 1; Figs. 1a, g–i, 4a; Table 2). In contrast to the ADn, when we examined all early-stage cases (Braak stages 0-II, n = 15 cases), we found that the other rostral thalamic nuclei, even those adjacent to the ADn, distinctly lacked ptau (Figs. 1a, 2a; Figure S2a; Table 2; Tables S3, S4). Among the postsynaptic targets of the ADn, both the EC and retrosplenial cortex (RS; BA 30, 29 and 26), but not the presubiculum (PreS), exhibited sparse ptau (EC score = 1, n = 3 cases, RS score = 0.5, n = 2 cases; Fig. 1b, j; Table S3). Mild tau pathology in CA2 (n = 2/3 cases) suggests the presence of four-repeat tau isoforms or primary age-related tauopathy (PART) in parallel with ptau in the ADn. At middle stages (Braak stages III-IV, n = 6 cases), the ADn contained moderate levels of ptau (coverage = 8.54%, ptau+ cells = 11.65 mm−2, intensity score = 2.65; Figs. 1c, g–i, 2; Figure S3a, b; Table 2; Tables S3, S4), whereas other rostral thalamic nuclei showed sparse ptau (AV: coverage = 0.15%, ptau+ cells = 14.49 mm−2, score = 0.5; paraventricular nucleus (PVT): coverage = 0.96%, ptau+ cells = 1.91 mm−2, score = 1.38; Fig. 1c, g–i; Tables S3, S4). In the cortex, ptau was observed at moderate-to-intense levels, with the RS, EC and prosubiculum (ProS) showing intense ptau immunoreactivity; the PreS displayed sparse immunolabeling at this stage (RS score = 3; EC score = 3; ProS score = 3; PreS score = 1; Fig. 1d, j; Table S3). In the RS, ptau was concentrated in the granular areas (BA 29 and 26). At late stages (Braak stages V–VI), which included cases with Alzheimer’s disease (Table 1), intense ptau was observed in the ADn, with a high density of ptau+ cells (n = 4/4 cases; coverage = 36.31%, ptau+ cells = 118.89 mm−2, intensity score = 3; Figs. 1e, g–i, 3; Tables S3, S4). The laterodorsal nucleus (LD) showed a high ptau density in the late stages, matching that of the ADn (Figs. 1i, 3; Table S3). The PVT had a similar ptau coverage to middle-stage ADn (n = 5/5 examined cases; 11.62%, score = 1.38; Figs. 1c, e, g, i, 3; Table 2; Tables S3, S4) and had the highest ptau+ cell count after the ADn (29.13 cells mm−2). The reuniens nuclear complex (RE) was similarly affected (score = 2). In contrast, the AV had lower coverage and ptau+ cell counts (coverage = 6.28%, ptau+ cells = 14.49 mm−2, intensity score = 1) followed by the mediodorsal nucleus (MD), which remained relatively sparse compared to the other examined nuclei (coverage = 1.62%, ptau+ cells = 2.84 mm−2, score = 1.25; Fig. 1e, g–i; Table 2; Tables S3, S4). The GABAergic thalamic reticular nucleus (TRN) lacked ptau immunopositive cell bodies; only axons and axon terminals were ptau+ (coverage = 2.98%, ptau+ cells = 0 mm−2, score = 2 Figs. 1e, g–i, 3; Tables S3, S4), consistent with previously published data. In the cortex, ptau severely affected each examined area (Fig. 1f, j; Table S3), consistent with previous studies. To provide additional context for the very early subcortical tau pathology in the ADn, we also examined the LC, which is susceptible to ptau at ‘pre-Braak’ stages. At early stages, in contrast to the sparse–moderate levels of ptau in the ADn, ptau expression in the LC was predominantly very weak and localized to processes resembling axons (n = 7/8 trace inclusions or sparse; n = 1/8 lacking ptau; n = Figure S4a, b; Table 2). In the middle and late stages, there was moderate ptau in somata, dendrites and axons (Figure S4c; Table 2). Interestingly, we detected neuromelanin in a few ptau+ somata. The DRn also showed moderate ptau in one tested Braak stage II case (Figure S4b). The lateral mammillary nucleus (LMB), which is presynaptic to the ADn (Fig. 1a), lacked ptau in n = 3/4 early-stage cases (Figure S5a; Table 2). In one Braak stage II case, the ADn showed moderate-dense ptau and the LMB showed mild–moderate ptau (Case 23, Figure S5b). In middle and late stages, moderate–dense ptau was present in the LMB (Figure S5c, d; Table 2). The adjacent medial mammillary nucleus (MMB) and tuberomammillary nucleus had a similar pattern to the LMB (Figure S5b; Table 2). The data largely confirm previous reports. In addition to neuronal ptau, we observed ptau+ ‘coiled bodies’ in the ADn (Figure S3c). Based on their size (~10 µm) and shape, we suggest that coiled bodies are localized to oligodendrocytes, which are typically overlooked in Alzheimer’s disease. We also detected ptau+ tufted astrocytes in six cases, which are associated with aging, Alzheimer disease, and other tauopathies. Despite widespread ptau in astrocytes of varying shapes, sizes, and locations, including within the ADn (Figures S3a, b, d, S5c), neuronal ptau was consistently detected within the ADn, presynaptic LMB, and postsynaptic RS, suggesting this pathway can develop tau pathology in parallel with aging-related tau astrogliopathy (ARTAG). These results reveal that distinct nuclei of the rostral thalamus are affected early on by ptau, with the ADn consistently having the highest ptau density and ptau+ cells across all stages (Figs. 1a, 4a; Figures S2a, S3a; Table 2; Tables S3, S4). Calretinin-expressing neurons accumulate ptau in the rostral thalamus We noticed that thalamic nuclei vulnerable to ptau were in CR-enriched regions (Figs. 2, 3, 4a) and hypothesized that CR+ neurons were sensitive to accumulating ptau. We performed double immunolabeling with CR and AT8 (for ptau) and observed colocalization in neurons within the ADn, PVT, and RE (Fig. 4a–d). In the TRN, CR-enriched neurons lacked ptau (Fig. 4f), consistent with the distinct lack of ptau+ TRN cell bodies (Fig. 1h). As the ADn contained ptau+ neurons even in control cases (Figs. 1a, h, 4a), we tested whether CR+ neurons were affected at early, middle and/or late stages. Even at Braak stage 0, CR was detected in the majority of ptau+ neurons (64.3% CR+ , n = 1 case; Fig. 4a, b, e). In the middle stage, a large proportion of ptau+ cells were CR+ (81.1%; n = 3 cases; Fig. 4c, e), and in the late stage, 71.1% were CR+ (n = 1 case; Fig. 4d, e). In conclusion, CR-expressing neurons were affected very early on, and at every stage, the majority of ptau immunopositive cells were CR+ in the ADn. Subcellular distribution of ptau in the anterodorsal thalamus After establishing that ADn neurons were especially vulnerable to ptau, we investigated the subcellular distribution of ptau to reveal how it spreads. To define synaptic structures at different stages of tau pathology, we examined ultrathin (~50–70 nm) sections of the ADn. We obtained electron microscopic samples from 4 cases (Cases 4, 12, 25, 17) that were appropriately preserved for quantitative analysis (Fig. 5a–d; Braak stages 0, II, III, and VI). We identified two main types of synaptic boutons with asymmetric synapses: large ~1–8 µm boutons (Fig. 5c; Figure S6a–c), consistent with presynaptic axon terminals from the mammillary body, and small <~1 µm diameter boutons (Fig. 5d; Figure S6a–c), resembling ‘classical’ cortical presynaptic terminals. Some presynaptic boutons from stage 0 (n = 20/70), from stage II (n = 3/106) and stage VI (n = 31/103) had a highly electron opaque (‘dark’) appearance, ranging from a homogeneous state to others with recognizable vesicles and mitochondria, but all showing collapsed, scalloped forms (Figure S2c). This may indicate degeneration of certain nerve terminals, and/or be a sign of selective vulnerability to post-mortem/fixation conditions; these terminals were omitted from our quantification. To identify subcellular ptau, we first examined cell bodies in the ADn, which contained abundant filaments (Fig. 5e). These resembled filaments previously found in the cortex of tauopathies including Alzheimer’s disease. We visualized ptau with silver-enhanced immunogold particles, and observed that ptau was specifically associated with the intracellular filaments (Fig. 5e, g; Figure S2g), thus unequivocally demonstrating the association of ptau with the originally described paired helical filaments at the ultrastructural level. Cell bodies also contained abundant lipofuscin (Fig. 4d, 5e). Filaments immunolabeled for ptau were also localized to dendrites (Fig. 5f; Figure S2d, g, h), and could be observed in large bundles (>1 µm) (Fig. 5g). Filament bundles were immunolabeled predominantly on the cytoplasmic surface, most likely due to reagents not penetrating into the bundle (Fig. 5g). We also detected ptau in myelinated axons (Fig. 5h). Given that ptau was localized to a variety of subcellular domains, we next investigated whether ptau can also be associated with axon terminals in the ADn. Subcortical vesicular transporter 2-expressing presynaptic terminals preferentially contain ptau Large presynaptic terminals of subcortical origin contain vesicular glutamate transporter 2 (vGLUT2). We observed strongly overlapping distributions of vGLUT2 and AT8 immunoreactivities at the light microscopic level, especially in the ADn, RE, PVT, and internal medullary lamina (Fig. 6a). The overlapping vGLUT2 and AT8 distributions suggested that vGLUT2 may be related to ptau. When we examined sections immunoreacted for both vGLUT2 and AT8, we discovered that ptau was localized to vGLUT2+ boutons (Fig. 6b, e, f; Figure S2b, d). Whereas some vGLUT2+ boutons showed no signs of abnormalities (Fig. 6c), others were degenerating (Fig. 6d; Figure S2c). The degenerating vGLUT2+ boutons had clumped mitochondria (Fig. 6d). Many of these boutons contained large (80–100 nm) double-walled vesicles (Fig. 6d), consistent with autophagy or the packaging and/or potential release of different forms of tau. Not all vGLUT2-positive degenerating boutons displayed detectable ptau, at least in the sections that we examined. In some degenerating boutons which were immunoreactive for both vGLUT2 and ptau, ptau+ bundles of filaments occupied a large proportion of the volume crowding out vesicles (Fig. 6b, e), which may cause impairments in neurotransmission. We also observed synaptic partners consisting of presynaptic vGLUT2+ boutons and postsynaptic dendrites that both contained ptau (Fig. 6f; Figure S2d), suggestive of transsynaptic spread between the mammillary bodies and ADn (Fig. 1a; Figure S5b). The size distribution of ptau-positive terminals also confirmed that large presynaptic terminals are preferentially affected by ptau pathology (Figure S6a–c). The distribution of presynaptic and postsynaptic ptau suggests transsynaptic spread Given the observation of ptau in both presynaptic terminals and postsynaptic dendrites (Fig. 5f, 6b, d–f; Figure S2b, d, g, h), we quantitatively characterized how the synaptic distribution of ptau changed across different Braak stages, examining 652 presynaptic boutons and postsynaptic dendrites, each of which was followed over several serial sections. At Braak stage 0, despite ptau being detectable at the light microscopic level (Figs. 1a, 4a), all sampled boutons and dendrites lacked ptau (n = 50/50 boutons, n = 62/62 dendrites; Case 4; Fig. 6g, h). Similarly, despite abundant ptau in the ADn at Braak stage II (Fig. 1i; Figure S2a; Table 2; Table S3), we did not detect ptau in boutons (n = 106/106) or dendrites (n = 109/109) at the electron microscopic level (Fig. 6g, h; Figure S2e, f), probably due to the limited sampling area. At Braak stage III, 5.8% of boutons (n = 6/104) and 21.6% of dendrites (n = 22/102) were ptau+ (Case 12; Fig. 6g, h; Figure S2d). In this stage, the proportion of synapses in which both the presynaptic boutons and the associated postsynaptic dendrites contained ptau was 3.9% (n = 4/104; Figure VGLUT2f S2d). At Braak stage IV, we detected ptau within dendritic appendages in close apposition to vGLUT2+ terminals, which also contained ptau (Figure S2b). At Braak stage VI, the proportion of affected boutons and dendrites greatly increased: 20.6% of boutons (n = 15/73) and 51.5% of dendrites (n = 52/101) contained ptau (Case 17; Fig. 6b, d–h). Furthermore, 12.3% (n = 9/73) of synapses consisted of both ptau+ boutons and ptau+ dendrites (Fig. 6f, g; Figure S2d). These data demonstrate that the proportions of both the presynaptic and postsynaptic elements containing ptau increase with Braak stage. Finally, we examined the relationship between presynaptic vGLUT2 and ptau across stages. At the early stage, we identified vGLUT2+ boutons (n = 18, Braak stage 0; n = 37, Braak stage II; Figures S2e, f, S6d), but did not detect ptau (Figure S6). But by the middle stage, from a total of 104 synaptic boutons, 5.8% (n = 6) were both vGLUT2 and ptau double immunopositive (Fig. 6i), whereas none of the vGLUT2 immunonegative boutons (n = 57) were ptau+. In other words, 100% of ptau+ boutons were vGLUT2+ (n = 6) and 12.8% of vGLUT2+ boutons (n = 47) were ptau+, supporting the hypothesis of selective vulnerability of subcortical vGLUT2+ synaptic terminals. Filamentous contacts with postsynaptic structures, known as puncta adherentia, are associated with mammillothalamic terminals. We identified puncta adherentia between vGLUT2+ boutons and postsynaptic dendrites containing ptau (Figure S2d). Moreover, the small corticothalamic boutons lacked ptau (Figure S6b, e), which indicates that ptau in the ADn is unlikely to have spread anterogradely from the cortex. In the late stage, out of a total of 73 synaptic terminals, an even higher proportion showed vGLUT2 and ptau colocalization (16.4%; n = 12; Fig. 6i), i.e., ~80% of ptau+ boutons (n = 15) were immunopositive for vGLUT2. And of all vGLUT2+ boutons (n = 38), 31.5% were ptau+. The data on the colocalization of vGLUT2 and ptau is even likely to be an underestimate, given that large ‘dark’ boutons are likely to be degenerating mammillothalamic terminals (Figures S2c, S6f), and we only sampled relatively few sections for each terminal. The above results suggest that vGLUT2+ boutons are strong candidates for the transsynaptic spread of ptau between postsynaptic ADn neurons and presynaptic mammillary body neurons within the Papez circuit (Fig. 7). | [
[
29457,
29460
],
[
33306,
33309
],
[
33903,
33908
],
[
33852,
33855
],
[
34267,
34270
],
[
34458,
34461
],
[
34617,
34620
],
[
34736,
34739
],
[
34959,
34962
],
[
34976,
34979
],
[
35037,
35040
],
[
35119,
35122
],
[
35185,
35188
],
[
35334,
35337
],
[
36139,
36142
],
[
36415,
36418
],
[
36649,
36652
],
[
37162,
37165
],
[
36496,
36499
],
[
37570,
37573
],
[
37591,
37594
],
[
37753,
37756
],
[
38278,
38281
],
[
38395,
38398
],
[
38483,
38486
],
[
38703,
38706
],
[
38347,
38350
],
[
38837,
38840
],
[
39249,
39252
],
[
39330,
39333
],
[
39454,
39457
],
[
39409,
39412
],
[
39262,
39265
],
[
39699,
39702
],
[
39588,
39591
],
[
39730,
39740
],
[
40079,
40082
],
[
40119,
40122
],
[
40211,
40214
],
[
40246,
40249
],
[
40830,
40833
],
[
42267,
42270
],
[
42335,
42338
],
[
42087,
42090
],
[
43420,
43423
],
[
44467,
44470
],
[
46565,
46598
],
[
46600,
46606
],
[
46659,
46665
],
[
46743,
46746
],
[
46815,
46821
],
[
46859,
46865
],
[
46939,
46945
],
[
46996,
47002
],
[
47058,
47064
],
[
47185,
47191
],
[
47416,
47419
],
[
47429,
47435
],
[
47600,
47606
],
[
47834,
47840
],
[
47995,
47998
],
[
48017,
48020
],
[
48879,
48882
],
[
49443,
49449
],
[
49544,
49550
],
[
50126,
50132
],
[
50191,
50197
],
[
50412,
50418
],
[
50482,
50488
],
[
50580,
50586
],
[
50613,
50619
],
[
50722,
50728
],
[
50917,
50923
],
[
51096,
51099
],
[
52365,
52371
],
[
52492,
52498
],
[
52511,
52517
],
[
52593,
52599
],
[
52855,
52861
],
[
52951,
52954
]
]
|
38447542 | intro | Introduction The World Health Organization (WHO) defines active aging as “the process of optimizing opportunities for health, participation and security in order to enhance quality of life as people age, allowing people to ‘realize their potential for physical, social and mental well-being throughout the life course.’” The list of contributors to healthy aging is long and varied with most items independent of hearing loss (HL) and balance disorders (BDs). Nevertheless, treating deafness and/or dizziness has proven positive effects on several health-related quality of life domains, which in turn benefits health status overall. HL and/or BDs are frequently linked to pathologies of the auditory or vestibular systems. These pathologies might appear in isolation or together. HL and/or BD represent a greater risk of cognitive impairment, frailty, and social exclusion with considerable negative consequences for their quality of life, that of those who care for them, and for the sustainability of health and care systems. The early detection of risks associated with aging (like HL and/or BD) using common telecommunications infrastructure approaches will enable an earlier intervention preventing their negative consequences. Age-related hearing loss (ARHL) is the loss of hearing that gradually occurs in most people as they grow older. It is one of the most common conditions affecting older and elderly adults. According to Roth et al., roughly 30% of men and 20% of women in Europe suffer from a hearing loss of 30 dB HL or more at age 70 and 55% of men and 45% of women at 80. About one-third of those who are affected at the European level have disabling hearing loss and it is estimated that around 900,000 have hearing loss severe enough to be candidates for a cochlear implant (CI). ARHL is a key communication disorder, which is characterized not only by a peripheral (cochlear) component but also by a central component. Severe hearing loss patients can barely understand spoken language. Even if they have an adequate tonal hearing sensation, they cannot understand complex acoustic stimuli, such as language or music, particularly in a noisy environment. The central neuronal processing speed and the timing of afferent integration is altered. Furthermore, a loss of inhibitory control and spatial memory has been observed as the result of the sensory (hair) cell loss and progressive deafferentation. Central ARHL must be regarded as an underrated factor, responsible for the break-down of interhuman communication in the elderly. This often leads to social isolation, abandonment of working life, and sub-depression. The lack of auditory information is also associated with cognitive dysfunctions and in extreme cases to age-related dementia. However, epidemiological studies show that the risk of developing a central ARHL is increased by 4–9% per year of age (beginning around 55) with increased prevalence in men. Taken together, this explains how hearing impairment negatively affects the quality of life of the elderly. Recent studies suggest that individuals with hearing loss are more likely to develop Alzheimer’s disease or other forms of dementia over time, and age-related hearing impairment is potentially a reversible risk factor for dementia and Alzheimer disease. The cause of BD in the elderly is usually related to multiple factors. Vestibular pathology is a common etiology, but their type varies: positional vertigo is very common, while Meniere’s disease or vestibular migraine have rarer presentation at that age but may already be present but undiagnosed in several subjects, which increases its prevalence. ARHL is usually the result of degenerative processes affecting the inner ear. However, there may be accompanying changes in the labyrinth at the cellular, microvascular, and metabolic levels, generating symptomatically permanent sensorineural hearing loss and/or vestibular disorders, leading to impaired balance. Hearing and balance impairments as well as falls are common among older people. Identifying modifiable risk factors for falls in older adults is of significant public health importance. While hearing is not typically considered a risk factor for falls, recent reports demonstrated a strong association between audiometric hearing loss and incident falls. Lin and Ferrucci report that hearing loss was significantly associated with the odds of reported falls. For every 10 dB increase in hearing loss, there was a 1.4-fold (95% CI, 1.3–1.5) increased odds of an individual reporting a fall over the preceding 12 months. Several mechanisms could explain the observed association between hearing loss and falls. There may be a concomitant dysfunction of both the cochlea and the vestibular organs given their shared location within the bony labyrinth of the inner ear. Decreased hearing sensitivity may also directly limit access to auditory cues that are needed for environmental awareness. Finally, the association of hearing loss with falls may be mediated through cognitive load and reduced attentional resources. Attentional resources are critical for maintaining postural control, and decrements in attentional and cognitive resources due to hearing loss may impair the maintenance of postural balance in real-world situations and increase the risk of falling. The latter two pathways, that suggest a possible causal pathway between hearing loss and falling, are intriguing because hearing loss is highly prevalent but remains vastly undertreated in older adults. To detect and typify hearing loss and BD among people older than 55 years of age and the epidemiological traits that might be linked to such disorders (primary prevention) To evaluate the impact of HL and/or BD on the domains that have impact on overall well-being and healthy aging in the elderly such as hearing ability, dependency, cognition, falls, and depression (in short: to evaluate the general status of quality of life in elderly individuals) To prove the positive impact of early intervention in HL and BDs among the elderly (secondary prevention) on quality of life, especially regarding their communication, cognitive, mental, and autonomy-related abilities To advocate for the care of hearing and balance disorders of elderly patients at a societal level among the general population The project “hearing and balance in healthy aging” is focused on the development of solutions that support active and healthy aging by enabling early detection and mitigation of risks associated with aging. Although the auditory thresholds of patients affected by ARHL have been studied, there are not enough data about the impact of HL on the quality of life of the elderly. There are few epidemiological data that research the link between ARHL and BD, diet, habits and physical exercise. Such information might be extremely useful to provide comprehensive, early care to elderly people affected by a HL linked or not to BDs, thereby contributing to existing prevention and treatment measures that reduce the impact of HL and/or BDs beyond the sensory field, on cognition, independence, and sociability. In addition, the information obtained might provide insights into the economic impact on society of HL and BDs in this population. The objectives of the project “H&B in healthy aging” are: In this article, the specific objective is the characterization of hearing, balance, and other associated disorders, in three population groups of individuals aged 55 and older. These groups are described below in the Population section. | [
[
4468,
4474
],
[
4489,
4495
],
[
4709,
4711
],
[
4475,
4478
],
[
5368,
5370
],
[
5063,
5065
],
[
4916,
4918
],
[
5601,
5607
],
[
5745,
5748
],
[
5760,
5765
],
[
5844,
5847
],
[
5859,
5864
],
[
6242,
6250
],
[
7219,
7222
],
[
5812,
5814
],
[
7481,
7484
],
[
7164,
7167
],
[
7026,
7029
],
[
5826,
5829
],
[
5516,
5519
],
[
7841,
7844
],
[
8327,
8333
],
[
9871,
9877
],
[
9901,
9904
],
[
10022,
10024
],
[
10331,
10333
],
[
10564,
10572
],
[
10869,
10877
],
[
11208,
11214
],
[
11532,
11534
],
[
11347,
11349
],
[
11229,
11231
],
[
10960,
10962
]
]
|
38259745 | results | Results Between October 1st, 2019, and May 30th, 2022, 225 patients benefited from neuropsychological evaluations or cognitive training and were included in the analyzes (103 Female, mean age 71.53 ± 15.36 years). The patient’s population consisted in post-traumatic disorders (8%), post-stroke patients (23%) and dementia patients (69%) (Alzheimer’s disease, Mild Cognitive Impairment, Frontotemporal Dementia, Lewy Body Dementia). A total of 13,958 treatments were included in the analyzes: 2,512 neuropsychological evaluations (18% of the total treatments) and 11,446 cognitive training (82% of the total treatments). Among these 13,958 treatments, 5,768 appointments were conducted in telemedicine and 8,190 in-person appointments were conducted in the outpatients’ clinic. During this period, the rate of telemedicine activity increased from 16.81% in January 2020 to 23.21% in May 2022. Peaks in telemedicine activity reached 85.64% in May 2020 and 83.65% in February 2021 (Table 1). In-person appointments in the outpatients’ clinic had a greater variability compared to telemedicine appointments, especially during the year 2020, as shown by the standard deviations analyzes (in-person appointments sd = 143.13 vs. telemedicine sd = 76.30, p = 0.017) (Figure 1). The monthly percentage change index for the outpatients (mean percentage variation = 13.83; 95% CI: −8.58, 42.67) and telemedicine activity series (mean percentage variation = 5.14%; 95% CI: −6.43, 16.46) during the COVID-19 pandemic, shows a positive trend concerning telemedicine continuity and utility. The bivariate spectral analysis yielded a significant common movement in the two series, with a significant peak involving a squared coherence of 0.421 (p = 0.032; phase angle, 3.03 radian) (see Figure 2), corresponding to a Fourier period of 5.2 months and with a 2.5-month lead relationship between the two time series. There was a significant positive correlation (r = 0.354, p = 0.038) between the number of telemedicine appointments and pandemic worsening expressed as the number of symptomatic patients hospitalized with SARS-CoV-2 in Italy (Figure 3). The first two peaks of SARS-CoV-2 contagion (March 2020 and November 2021) were immediately followed by an increase in telemedicine appointments. Strikingly, patients were faster to switch from in-person to telemedicine appointments at the third worsening of the Italian pandemic situation. This was reflected by the rise in telemedicine activity preceding the third peak of the pandemic, in Italy (Figure 4). | [
[
8563,
8571
],
[
8729,
8736
],
[
8808,
8816
],
[
8836,
8844
],
[
12213,
12221
],
[
12240,
12250
],
[
12398,
12408
],
[
12533,
12541
]
]
|
33101276 | abstract | The R47H variant in the microglial triggering receptor expressed on myeloid cell 2 (TREM2) receptor is a strong risk factor for Alzheimer's disease (AD). To characterize processes affected by R47H, we performed an integrative network analysis of genes expressed in brains of AD patients with R47H, sporadic AD without the variant, and patients with polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy (PLOSL), systemic disease with early-onset dementia caused by loss-of-function mutations in TREM2 or its adaptor TYRO protein tyrosine kinase-binding protein (TYROBP). Although sporadic AD had few perturbed microglial and immune genes, TREM2 R47H AD demonstrated upregulation of interferon type I response and pro-inflammatory cytokines accompanied by induction of NKG2D stress ligands. In contrast, PLOSL had distinct sets of highly perturbed immune and microglial genes that included inflammatory mediators, immune signaling, cell adhesion, and phagocytosis. TREM2 knockout (KO) in THP1, a human myeloid cell line that constitutively expresses the TREM2- TYROBP receptor, inhibited response to the viral RNA mimetic poly(I:C) and phagocytosis of amyloid-beta oligomers; overexpression of ectopic TREM2 restored these functions. Compared with wild-type protein, R47H TREM2 had a higher stimulatory effect on the interferon type I response signature. Our findings point to a role of the TREM2 receptor in the control of the interferon type I response in myeloid cells and provide insight regarding the contribution of R47H TREM2 to AD pathology. | [
[
115,
119
],
[
195,
200
],
[
239,
258
],
[
260,
262
],
[
303,
307
],
[
386,
388
],
[
389,
397
],
[
403,
407
],
[
418,
420
],
[
446,
454
],
[
460,
536
],
[
538,
543
],
[
546,
562
],
[
580,
588
],
[
629,
634
],
[
650,
694
],
[
696,
702
],
[
723,
725
],
[
773,
778
],
[
779,
783
],
[
784,
786
],
[
851,
873
],
[
902,
907
],
[
937,
942
],
[
1023,
1035
],
[
1098,
1103
],
[
1121,
1125
],
[
1129,
1134
],
[
1187,
1192
],
[
1194,
1200
],
[
1255,
1264
],
[
1335,
1340
],
[
1400,
1404
],
[
1405,
1410
],
[
1524,
1529
],
[
1655,
1659
],
[
1660,
1665
],
[
1669,
1671
]
]
|
38594369 | title | Design and methods of the mobile assessment of cognition, environment, and sleep (MACES) feasibility study in newly diagnosed breast cancer patients | [
[
82,
87
]
]
|
39072940 | intro | INTRODUCTION Alzheimer's disease (AD) is a neurodegenerative disease affecting the brain with high incidence in the elderly. The typical clinical manifestations include progressive impairment or loss of cognitive function. Senile plaques formed by extracellular β‐amyloid (Aβ) deposition, and neurofibrillary tangles formed by intracellular hyper‐phosphorylated tau protein aggregation in neuron are two hallmark pathological changes in the brains of AD patients, and they also represent key factors inducing or mediating cognitive damage. AD is the main cause of dementia, accounting for about 70% of the total cases of dementia. With the intensification of the aging of the world population, especially in the elderly over 75 years old, the incidence of AD increases year by year, and, as a matter of fact, has become a dire global public health problem. According to the World Alzheimer Report 2019, there are as many as 50 million AD patients worldwide and the number is expected to double every 20 years and increase to 152 millions by 2050, which not only brings heavy economic and emotional burdens to the society and families, but also poses a great challenge to the national medical and healthcare systems. It has been more than a century since AD was first reported, but there is still lack of effective measures to modify or prevent the disease progression, albeit drugs existing to alleviate the symptoms. Research efforts in the past decades have greatly increased the understanding of the clinical histopathology of AD, particularly neuronal cell death in different brain regions and impairments in cognitive and neurological functions. At the same time, with the multiple effects of genetic or environmental factors, people are also increasingly aware of the extreme complexity of AD etiology. In terms of etiological classification, except for about 5% of familial hereditary AD cases, the rest are late‐onset or sporadic AD. Currently, the Aβ cascade hypothesis, tau protein hypothesis, and cholinergic hypothesis have been proposed for the pathogenesis of sporadic AD, but there is still no definitive conclusion on what mechanism dominates the occurrence and development of AD. Neuroinflammation, oxidative stress, mitochondrial dysfunction and autophagy lysosome system deficiency are considered to be the biological pathways that affect the pathology of AD. The disorder of any one pathway may trigger a vicious cycle of cell damage and neurodegeneration, which plays a crucial role in the development of AD. In particular, oxidative stress, which is closely related to AD and aging, can be targeted by antioxidants, which provides a possible direction for the treatment of AD. Reactive oxygen species (ROS) are a group of oxygen‐derived chemicals with short life but high activity. They are usually produced at a moderate level and eliminated in time through a variety of antioxidant mechanisms under physiological conditions. ROS as a signal molecule play an important role in maintaining cellular and tissue homeostasis. The increased production of free radicals can lead to oxidative stress, if the endogenous antioxidant capacity of the cell is insufficient to offset the production of these free radicals. The brain is vulnerable to oxidative stress due to its high oxygen consumption, relatively high level of polyunsaturated fatty acids that are sensitive to oxidation and relatively low level of antioxidant enzymes. Oxidative stress is considered to be the main cause of AD progression. Severe neuronal oxidative damage have been observed in the atrophic brain tissues of AD patients. Chronic accumulation of AD‐related abnormal or misfolded proteins are well‐documented to induce excessive production of ROS, impair the antioxidant capacity of cell, or both, resulting in increased ROS levels and oxidative stress and thus changes in normal functions of cells in the brain. Aging is one of the main reasons for the gradual decline of brain functions, which can lead to progressive cognitive impairment. It is also one of the pathological risk factors of AD. During the process of aging, the increase in free radicals can damage the structure and function of the brain. It was shown that the production of ROS is related to synaptic dysfunction and tau protein phosphorylation. The high level of ROS in the brain can result in the oxidation of proteins, lipids, and DNA, and eventually lead to neuronal loss and dysfunction. Oxidative stress can also drive the progress of tau pathology in many ways, such as producing fatty acids, stimulating fibrosis tau protein aggregation, activating glycogen synthase kinase‐3 (GSK‐3β), and increasing hyperphosphorylation of tau proteins at multiple sites. Under normal circumstances, endogenous antioxidants, such as trace elements (zinc and selenium), vitamins (vitamin C and vitamin E), polyphenols and coenzyme Q10, and antioxidant enzymes, including superoxide dismutase (SOD), catalase, and glutathione oxidoreductase, can resist or prevent oxidative damage. However, studies have reported that the level of antioxidant enzymes in the brains of patients with AD and mild cognitive impairment is significantly reduced, and the brain becomes more vulnerable to oxidative damage. Therefore, the use of antioxidants has been considered as a promising therapeutic strategy to alleviate oxidative stress‐induced or mediated neuronal damage and loss of function. Previous studies shown that long‐term administration of D‐(+)‐galactose in rodents can produce aging‐related changes that are similar to those observed in human beings, including the formation of excessive free radicals and the reduction of antioxidant enzyme activity. Therefore, D‐(+)‐galactose‐treated mice or rats have become a widely used experimental model for researching aging and underlying mechanisms. Melatonin is a neurohormone secreted by the pineal gland of the brain that regulates the circadian rhythm and plays an important role in restricting neuroinflammation and oxidative damage. Studies have shown that the melatonin level in AD patients is lower compared with that in normal subjects. As shown in vivo and in vitro, melatonin supplementation effectively inhibited Aβ deposition and tau protein phosphorylation and, furthermore, treatment with melatonin restored cognitive function in AD transgenic mice. Mitochondria are the main source of ROS in cells and also the target for oxidative stress. Coenzyme Q10 is a vitamin‐like substance located in the inner membrane of mitochondria and acting as an electron carrier in the mitochondrial respiratory chain and has a redox capacity. Coenzyme Q10 is one of the most powerful endogenous membrane antioxidants and has been shown to protect neuronal cells from oxidative stress damage in vivo and in vitro. Moreover, coenzyme Q10 can reduce malondialdehyde (MDA) expression, regulate SOD activity, reduce Aβ plaque formation, protect AD‐related loss of synaptic plasticity and improve cognitive function. Lecithin is rich in choline and can combine with acetyl coenzyme A to form acetylcholine. Lecithin has been shown to reduce Aβ‐induced neuronal damage. Lecithin alone can also increase the expression of neurotrophic factors in the brain of mice, enhance the antioxidant capacity, and improve learning and memory. Therefore, the antioxidant therapy is considered to be an effective strategy to prevent or alleviate the pathological progress of oxidative neuronal damage and neurodegenerative conditions such as AD. The clinical trial results of using different nutritional supplements, singly or in combination in patients with AD, are different. As aforementioned, the pathogenesis of AD is highly complex, and the use of a single drug in preventing and treating AD is often not effective, leading to the notion that combined treatment with multiple drugs have a better or more significant effect. Therefore, in this study, we selected melatonin, coenzyme Q10, and lecithin, three widely used nutrients with an antioxidant capacity, and examined the neuroprotective effects of applicating them individually or in combination on oxidative stress‐induced or mediated neuronal loss, and D‐(+)‐galactose‐induced neuronal damage and cognitive impairment, and the mechanisms of their actions. | [
[
2927,
2935
],
[
3415,
3423
],
[
2835,
2838
],
[
4214,
4220
],
[
4465,
4468
],
[
6094,
6102
],
[
7620,
7628
],
[
8096,
8101
],
[
8252,
8256
],
[
8260,
8264
],
[
6773,
6776
],
[
6997,
7000
],
[
7077,
7080
],
[
7194,
7197
],
[
7452,
7460
],
[
8599,
8607
],
[
8861,
8876
],
[
9773,
9777
],
[
10146,
10154
],
[
8756,
8759
]
]
|
39131307 | abstract | Several age-related oral health problems have been associated with neurodegenerative diseases such as Alzheimer’s Disease (AD), yet how oromotor dysfunction in healthy aging differ from those found in pathological aging is still unknown. This is partly because changes in the cortical and biomechanical (“neuromechanical”) control of oromotor behavior in healthy aging are poorly understood. To this end, we investigated the natural feeding behavior of young and aged rhesus macaques (Macaca mulatta) to understand the age-related differences in tongue and jaw kinematics. We tracked tongue and jaw movements in 3D using high-resolution biplanar videoradiography and X-ray Reconstruction of Moving Morphology (XROMM). Older subjects exhibited a reduced stereotypy in tongue movements during chews and a greater lag in tongue movements relative to jaw movements compared to younger subjects. Overall, our findings reveal age-related changes in tongue and jaw kinematics, which may indicate impaired tongue-jaw coordination. Our results have important implications for the discovery of potential neuromechanical biomarkers for early diagnosis of AD. | [
[
567,
582
],
[
584,
598
]
]
|
38590779 | title | Role of autophagy in ischemic stroke: insights from animal models and preliminary evidence in the human disease | [
[
98,
103
]
]
|
36348357 | title | ApoE in Alzheimer's disease: pathophysiology and therapeutic strategies. | [
[
0,
4
],
[
8,
27
]
]
|
39290562 | methods | Methods The systematic review was undertaken in accordance with the PRISMA criteria to ensure the reliability and comprehensiveness of the data and findings. The English databases searched included PubMed, Embase, the Cochrane Library, and the Web of Science and Scopus. The language used was English, according to the authors’ level of comprehension. Literature search strategy An electronic search was conducted in the databases with free terms, subject heading terms and key words. The keywords are as follows, (“Probiotic” OR “Gut microbiota” OR “Microbiome”) AND (“Alcohol addiction” OR “Alcohol use disorder” OR “Alcoholism”) AND (“Probiotic intervention” OR “Probiotic therapy”) AND (“Alcohol dependence” OR “Alcohol craving”) AND (“Review article” OR “Clinical trials”) AND (“Efficacy of probiotics” OR “Probiotic treatment for alcohol addiction”) AND (“Substance use disorder” OR “Alcohol withdrawal”) AND (“Evidence synthesis” OR “Patient outcomes”). Studies that provide data on the effects of probiotics in the context of alcohol addiction published from January 2012 to November 2023 were included in this review. This period was selected to ensure the inclusion of the most up-to-date and relevant studies, reflecting the latest advancements and current trends in this rapidly evolving field. Preliminary searches indicated that older studies did not significantly contribute new knowledge or insights relevant to the objectives of this study. Table 1 presents the specific inclusion and exclusion criteria based on the PICOS (i.e., population, interventions, comparisons, outcomes, and study design) strategy. Study selection Prior to conducting the database search, all researchers underwent a comprehensive training session to ensure a uniformity of ideas and a thorough understanding of the inclusion and exclusion criteria. The databases were thoroughly searched by the researchers, who then independently assessed the studies. The initial screening of the studies involved evaluating their titles and abstracts, followed by a thorough examination of the complete texts of the selected studies to determine their eligibility based on the predefined inclusion and exclusion criteria. Ultimately, the researcher conducted a comparison of the screened full-texts. Divergence and discrepancies were effectively addressed by a process of dialogue and vote among the researchers. The data extraction process involved gathering information from eligible studies using a predetermined template. This template encompassed several aspects such as the study’s title, authors, publication year, research design, participant characteristics, intervention details, outcome measures, results and statistical findings, as well as the conclusions and implications drawn from the study. The assessment of the methodological quality of the studies (Figure 1) included in the analysis was conducted using suitable techniques, such as the Cochrane Risk of Bias tool for clinical trials. | [
[
8639,
8646
],
[
11579,
11590
]
]
|
38813371 | intro | Introduction Mild cognitive impairment is a prodromal state of dementia that manifests with subtle balance control and gait deficits which may affect activities of daily living and contribute to the two folded increased risk of falls in this population. Consequences of these falls lead to reduced physical functioning, dependency, and long-term disability, ultimately increasing the likelihood to develop dementia or Alzheimer's disease. A majority of falls occur during activities of daily living that involve performing a motor and cognitive task simultaneously (i.e., dual tasking). Therefore, understanding the biomechanical basis of balance control deficits under dual task conditions may inform the development of fall prevention strategies for this population. Studies have used dual task methods to evaluate how individuals allocate cognitive/motor resources when simultaneously attending to two tasks. If the attentional demands are greater than the capacity of an individual, dual tasking results in cognitive-motor interference, such that there is deterioration of performance on either or both tasks. Previous studies have shown that older adults with mild cognitive impairment (OAwMCI) experience higher cognitive-motor interference than cognitively intact older adults (CIOA) during volitional balance control tasks, resulting in increased standing postural sway and reduced gait speed while performing a cognitive task (e.g., visual search, digit span recall, word recall). However, activities of daily living do not only involve gait and working memory tasks, but also entail the ability to recover from unpredictable balance threats induced by the environment (i.e., reactive balance control). Understanding biomechanical factors contributing to cognitive-motor interference in OAwMCI compared to their healthy counterparts remains to be explored. When an unexpected balance loss occurs, the CNS recruits feedback mechanisms to respond to the balance loss via compensatory strategies, which are modified online based on the perceived perturbation magnitude. In case of small magnitude perturbations, in-place ankle or hip strategies are recruited and as the magnitude becomes larger, a change in support strategy via stepping or grasping becomes necessary to recover postural stability (i.e., position and velocity of the COM relative to the displaced base of support). While such responses are triggered either by the short (spinal segmental) or long-loop (brain stem) reflexes, it is postulated that higher-cortical centers further relay the perturbation-specific sensory information (i.e., perturbation displacement, acceleration, and velocity) to optimize postural responses via the transcortical loop. Sensorimotor decline related to healthy aging can delay the ability to perceive and integrate sensorimotor information to initiate stepping, resulting in increased number of compensatory steps, delayed step initiation, and increased limb collisions in response to large intensity perturbations compared to young adults. There is limited understanding whether a state of mild cognitive impairment could further reduce the ability to recover from unexpected external perturbations. Our previous findings indicate that OAwMCI exhibit deteriorated reactive balance control compared to young adults and CIOA, including delayed step initiation time, reduced step length, and reduced postural stability when exposed to large magnitude perturbations. Further, OAwMCI were unable to modulate postural responses at higher perturbation intensities. These deficits are potentially attributable to the structural and functional cortical impairments observed in OAwMCI, as there is preliminary evidence that dual task reactive responses are worse among people with cortical lesions (e.g., stroke, Parkinson's disease, concussion injury). However, these responses were observed in a controlled environment where the individual had nothing but the motor task (i.e., slip-like perturbations) to attend to. Real-life environments may incorporate the additional dynamics of perceptual cognitive demands, like standing and visually searching for cues, or standing and tracing items in the environment. Research has shown that OAwMCI experience a decline in visual processing capacity, visual search, and attention-related processing, which leads to difficulty processing moving visual scenes during standing, causing losses of balance. These scenarios may require more substantial attentional demands to recruit the appropriate motor strategies to recover from unexpected balance control threats, due to potential overlapping resources between reactive balance control and cognitive function. In line with this, our recent study in young adults showed that reactive postural stability in response to large magnitude perturbations was significantly lower while performing visuomotor games than during single task reactive balance. There is no study to date that has examined reactive responses under dual task conditions challenging the perceptual cognitive function to understand the pattern of cognitive-motor interference in a real-life like environment in older adults with and without cognitive impairment. For this reason, this study primarily aims to determine the differences in single task and dual task reactive responses between CIOA and OAwMCI while performing two different perceptual cognitive tasks. We first hypothesize that OAwMCI will have reduced reactive balance control [indicated by reduced margin of stability (MOS), increased number of falls] compared to CIOA in both single and dual tasking. Secondly, we hypothesize that OAwMCI will show higher performance errors on cognitive tasks compared to CIOA during single and dual tasking. Lastly, we hypothesize that OAwMCI will demonstrate higher cognitive-motor interference than CIOA (i.e., greater reduction in performance in dual task vs. single task), due to difficulties allocating attentional resources in challenging conditions. Due to the constantly changing nature of cognitive demands in real-life tasks, examining the ability to recover from unexpected balance loss under perceptually challenging conditions could help understand the influence of cognitive function on reactive balance control. | [
[
6331,
6337
],
[
8105,
8108
]
]
|
38079409 | results | Findings The findings present a narrative and charting of the data from the 45 papers that met the review criteria and this data is mapped onto the five objectives outlined in step one of the framework Within this review, interventions were grouped into five broad types: nature (n = 6 papers), memory/cognitive (n = 11 papers), social (n = 17 papers), animal (n = 4 papers), or physical/sensory (n = 7 papers) based interventions. Types and range of psychosocial interventions used in day care service Several different types of psychosocial interventions were identified as a single or combined interventions (Table 2). Nature based interventions representing the person’s experience or understanding of their own natural world included: green care farm, farm based garden based meaning in life in general.Memory/cognitive-based interventions related to reminiscence structured sessions, life story interviews, life review programme, life story work, emotional therapy and cognitive behavioural therapy Combined approaches included combined cognitive training, music and art therapy combined environmental barriers and cognitive behavioural therapy and combined reality orientation and cognitive behavioural therapies. The social based interventions related to a single or range of activity combinations. Single social activities related to board games comedy workshops humanoid robot, adult care programme, technological group activity and art activities Combined social interventions were utilised in five studies and within two of these studies occupational activities were incorporated In addition, the broad activity of music was identified and included song writing listening to music to stimulate memory ‘singing for the brain’ and the development of a music application prototype to assist in human-computer interaction Animal-based interventions related to the use of animal assist (canine) therapy and an equine-assisted therapy intervention. Physical/sensory-based interventions evident within this review included multisensory environments, aromatherapy dance activity, exercise programmes., while Rylatt utilised a creative therapy exercise programme incorporating a combination of interventions (dance, drama, music, movement). Use and facilitation of psychosocial interventions Within the facilitation of psychosocial interventions several were ongoing while others are facilitated within a specific period/time limited specific programmes. Each study within the five broad types: nature, memory/cognitive, social, animal, or physical/sensory based interventions are presented in Table 3. Participants attended a green care farm and also a non-time limited programme with a similar structure for each day across all sites. Whereas programmes focused on the use of GCFs to promote physical activities and the use of a six-week programme of weekly gardening session. Memory/cognitive-based interventions that were time limited ranged from five days, two weeks, four weeks, six weeks, eight weeks and 12 weeks. Other programmes ran for over 12months, over two years or were open-ended. essions ranged from one hour/day, three hours a day, one session per week, two/three sessions per week, four sessions per week, six/eight sessions per week, ten sessions per week to thirty-three sessions per programme. Facilitation was led by the primary researcher or a facilitator but involved the inclusion of research assistants, staff or another facilitator. Social based interventions were individual sessions or group sessions The group programmes were open to whomever attended the centre on the day however, in one study six people participated in the original programme and ten in the modified programme. Programmes ran for four weeks, six weeks eight weeks, three months or appeared to be open-ended. Sessions ran for 45-60minutes, one hour/day, two and half hours/day, two/three hours per week or twice/week. Two papers do not describe their use or facilitation while in other papers facilitation was supported by staff and therapists. Animal-based interventions were time limited ranging from one to two session per week over a period of four weeks, 12 weeks or 12 months. Professional animal handlers were involved in all four programmes in conjunction with other facilitators. Physical/sensory-based interventions were time limited ranging from four weeks two months, four months. In contrast another review reported time limits from 2 weeks to 12months while another two exercise programmes were not time limited. Sessions lasted up to 20 minutes, 30 minutes 40 minutes to one hour/day. Reported evaluation methods A range of evaluations methods were used for the psychosocial interventions and can be grouped under into four categories 1) instrument/measurement measures (a broad spectrum of bio-psycho-social intervention measures) 2) exercise/functional measures, 3) qualitative measures and 4) activity measures and mapped under the five broad types: nature-based memory/cognitive, social, animal, or physical/sensory based interventions identified in this review (Table 4). Five papers did not report outcome measures. Within the first category (instrument/measurement) fifty outcome measures were identified across 26 papers. The mini-mental state examination was reported in eight papers. the Clinical Dementia Rating Scale in five papers. The Geriatric Depression Scale and the Cornell Scale for Depression in Dementia in four papers. The Cohen-Mansfield Agitation, the Frontal Assessment Battery the Montreal Cognitive Assessment, the Neuropsychiatric Inventory and the Quality of Life in Late-Stage Dementia Scale , in two papers. The remaining measure were reported in one paper each: Agitated Behavior Scale, Algase Scale, Alzheimer’s Disease Assessment Scale-Cognitive subscale, An ethogram catalogue, Apathy Inventory Scale, Apathy Scale for Institutionalized Patients with Dementia Nursing Home version, Barthel Index, Beck Anxiety Inventory, Behavior Intervention Monthly Flow Record, Behavioral Observation Recording, Berg Balance Score, Caregiver Exit Survey, Caregivers’ Survey, Clinical Global Impression-Improvement, Consortium to Establish a Registry for Alzheimer’s Disease ‐ Neuropsychological Battery, Daily Observation Scale, Digit Span task ‐ Finnish Wechsler Memory Scale, Emotional Satisfaction Index, Global Deterioration Scale, Instrumental Activities of Daily Living, Interview for Deterioration in Daily living in Dementia, Korean Dementia Screening Questionnaire-Cognition, Korean version ‐ Mini-Mental Status Examination, Lawton Scale of Activities of Daily Living, Milan Overall Dementia Assessment, Modified Nursing Home Behavior Problem Scale, Modified Philadelphia Geriatric Center Affect Rating Scale, Neuropsychiatric Inventory Nursing Home, Psychotic Behavior Assessment Record, Quality of Life in Alzheimer’s Disease, Quality of Life in Dementia Scale, Recreational Activities of Daily Living Scale, Salivary cortisol, Sedentary behaviour questionnaire, Seoul Neuropsychological Screening Battery, Seoul-Instrumental Activities of Daily Living, Severe Mini Mental State Examination, Short Form Health Survey, Social Functioning in Dementia Scale, the Clock Drawing Test and the Trail Making Test. Within the second category (exercise/functional) eight outcome measures were identified across four studies and all outcome measures were reported in one paper each; Accelerometer, Actigraphs, Barthel Index, Functional Ability Tests, Inclinometer, Interview for Deterioration in Daily living in Dementia, Non-exercise equation to calculate cardio-respiratory fitness and the Timed Up and Go-test. The third category (qualitative) highlighted five outcome measures across twenty papers. Interviews were reported in 11 papers. Observations were reported in eight papers.Video recording in five papers. Reflective discussions in two papers and photograph in one paper.The final category (activity) highlighted three outcome measures across three papers. Outcome measures identified were an activity observation form, verbal fluency and a self-designed tool. Reported outcomes A broad range of outcomes were identified within the papers reviewed. However, four papers identified no real difference and overall low to medium effect or maintenance post intervention was generally reported. These can be grouped under into four categories 1) increases in functioning, 2) social, 3) health and well-being and 4) enablement outcomes. Increases in functioning was reported in 16 papers across four elements: cognition physical activity/ability, activities of daily living and social skills. Increases in social functioning was reported in 24 papers across eight elements. Connection/engaging with others was evident in 11 papers. Increased communication was evident in seven papers. Building relationships was evident in five papers. Increased participation and increased social interaction were evident in four papers each. While increased enjoyment was evident in three papers and increased group connection and knowing the person were evident in one paper each. Health and well-being were reported in 17 papers across six elements and the main element was reduction in behaviours of concern Increased mood was evident in five papers.Three elements were all evident in three papers each: emotional well-being, increased well-being, quality of life. The remaining element depression was evident in two papers. Enablement was reported in seven papers across ten elements, two elements motivation and a sense of belonging were reported in two papers each. The remaining elements were all report in a single paper: a sense of accomplishment, increased agency, increased choice, collaborations, increased confidence, inclusion, a sense of freedom and being self-conscious. Reported adaptations Of the studies reviewed 19 reported on adaptations or adjustments made when implementing the psychosocial interventions. This provided important contextual factors around implementation. Five studies within the nature-based interventions grouping are relevant here. Ibsen and Eiksen recounted how all the interviews took place at the farms so that participants could better remember or relate to the day care setting. While Noone and Jenkin reported that over the course of the gardening project, the researcher developed a relationship with each participant which enabled participants to feel comfortable communicating their level of willingness to participate in a particular activity. Chang et al. described the focus of each reminiscence session being on a particular life phase with objects relevant to that period introduced for discussion and incorporating the use of records and interviews about older people’s early lives and interests. Gregory adjusted the timings of the intervention depending on factors such as the participant’s ability to concentrate and communicate and the intrusion of others into the space. The participants’ words were recorded and shown to them to emphasise how effectively they had been able to communicate their life stories and in read back sessions the poet read the participants’ poem aloud. Finished poems were sent to GP and kept in patient files. Lin et al. revised the life review programme to compress it into a short format that was administered in 10 successive sessions over 2 weeks. Within the memory/cognitive-based interventions grouping two papers are relevant. Jung et al. reported that the main activity stage comprised activities for strengthening memory and management function and for increasing attention, concentration, and space-time, perception, concept formation and reasoning, composition, language, and computational abilities. Help was provided if required by the participant and initially, an easy level programme was gradually adjusted to appropriate and slightly difficult levels executed to develop person interest, sense of achievement and confidence. The programme differentiates from existing chair-based exercises by including bilaterally asymmetric activities that involves the left side of the body moving differently to the right side of the body as well as activities requiring mirroring a partner to rhythmic music that evokes memories and reminiscence. Kallio et al. tailored training according to the participants’ cognitive abilities and it was implemented either in small groups of two to four persons, or individually when needed due to difficulties in concentration, or lack of a training pair. The difficulty level was tailored during the sessions, but it was not automatic as in computerized training. Six papers within the social-based interventions grouping six papers reported adaptations or contextual factors that determined how the intervention was adjusted. Gjernes described the staff member initiating discussions and serving as the engine that made the network social. Typically, the staff member used strategies to involve every person in the knitting activity. The knitters were supported to remember and participate in telling their own stories. The staff member was both a potential helper and a member of the social network. The staff could provide knowledge and skills when needed or could assist the knitters in their problem-solving efforts. Lancioni et al. described how in the modified program version the computer presented photos and videos, providing encouragements to talk as well as attention and guidance. In another individualised music playlists were developed by asking caregivers about the participant’s favourite music or by playing different songs for participants to see their reactions. During the study, some participants listened to their favourite songs repeatedly, others listened to a variety of songs. Distraction was minimised by closing the door and ensuring the participant’s physical comfort. Peeters et al. reported how the researcher used the analogy of a house with different rooms, in which buttons represent doors to move between rooms to explain the navigation through the different screens of music collections management functionality. Hattford-Letchfield reported how a comedy workshops did not work to an exact script but allowed scenarios to develop based on the main theme. Experiential drama techniques were used to work with the issues identified by the participants. A selection of photographs was made into a ‘scrapbook’ of the project as a whole that could be used as a reminiscence tool after the project had finished. Cheung et al. adapted the Play Intervention for Dementia and integrated elements of cognitive stimulation in six identified mind–body functional domains and followed the principles of cognitive stimulation. During the sessions, participants could exercise their creativity in a cheerful and respectful environment, without anyone judging their (dis-)ability or without pre-set rules. Within the animal-based interventions grouping two papers reported adaptations. Both papers described how sessions were designed to follow a protocol but could be individually tailored to each participant based on the care workers’ knowledge of the participant. Hence, none of the animal assisted activities were mandatory and the sessions also included naturally occurring activities between the participants as well as between each participant and the dog. Within the physical/sensory-based interventions grouping two papers reported adaptations. Karania71 describes how this programme differentiates from existing chair-based exercises by including concurrent bilaterally asymmetric activities that involves the left side of the body moving differently to the right side of the body, as well as activities requiring mirroring a partner to the sound of rhythmic music that evokes memories and reminiscence. Chang, et al. described a bespoke exercise programme designed as a series of exercise training interventions aimed at maintaining activities of daily living abilities with adaptations made to the comedy workshops and dance activities. Aguinaga and Marquez reported modifying their dance intervention where participants wore an orange Velcro bracelet on their right wrist and a green Velcro bracelet on their left wrist to help them distinguish between moves to the left and right. The programme was adapted as needed by revising the dance moves in ways that still challenged participants physically and cognitively but did not overwhelm them or put their safety at risk. | [
[
15490,
15495
],
[
15581,
15587
],
[
15604,
15610
],
[
34944,
34956
],
[
35380,
35383
],
[
35972,
35978
],
[
36159,
36162
],
[
36064,
36067
],
[
42946,
42954
],
[
44305,
44310
],
[
47253,
47265
],
[
47452,
47463
],
[
47478,
47490
],
[
47815,
47821
],
[
47935,
47946
],
[
48038,
48050
],
[
48218,
48230
],
[
48289,
48296
],
[
48843,
48854
],
[
49395,
49407
],
[
49492,
49499
],
[
50622,
50633
],
[
50688,
50700
],
[
50748,
50760
],
[
50909,
50920
],
[
51418,
51430
],
[
51821,
51833
],
[
52170,
52181
],
[
52228,
52239
],
[
52376,
52388
],
[
52413,
52424
],
[
52433,
52436
],
[
53194,
53206
],
[
53463,
53475
]
]
|
38334239 | intro | INTRODUCTION Parkinson's disease (PD) is a multifaceted neurodegenerative condition impacting upon many aspects of an individual's physical and psychological well‐being. PD is the second most common neurodegenerative disease, and the most common neurodegenerative movement disorder worldwide, with prevalence tending to increase with age. It is expected that by 2040, the global incidence of PD will exceed 12 million. The three main physical symptoms of PD are tremors, muscle stiffness, and slowness of movement. In addition, PD can impact an individual's mental well‐being, with symptoms of depression, delusions, paranoia, hallucinations, and PD‐associated dementia commonplace resulting directly from PD itself or through medication side effects. It is well documented that individuals with PD present with higher incidences of mental health (MH) problems such as depression, anxiety, schizophrenia, and psychotic symptoms when compared to the general population. More specifically, up to 40% of people with PD (PwPD) will have depression or anxiety, whereas this figure is only 17% in the general population. Current NICE guidelines do not address or provide specific recommendations for MH problems in this population, instead referring to existing generic guidelines on depression in adults with chronic health problems and how to access allied health professionals (e.g., physiotherapists and PD nurse specialists). This is a striking contrast to other neurological conditions such as multiple sclerosis, where guidelines include specific recommendations for regular cognitive, emotional, or MH screening. Evidence in older populations suggests a relationship between physical and psychological presentations. There is, however, a paucity of evidence to substantiate such a relationship in the PD population. From available evidence, it has been suggested that PwPD feel that anxiety may amplify their physical symptoms, and when they become more anxious the incidences of freezing of gait increase. To add to this, a number of studies have suggested that as anxiety increases, so does the severity of motor symptoms as assessed using the Unified Parkinson's Disease Rating Scale (UPDRS). Although these studies show that there might be an association between physical function and psychological symptoms, it is not well understood. It should also be noted that the UPDRS is a global measure of PD severity and includes both motor and non‐motor domains. This relationship is yet to be confirmed in more specific measures of physical function such as balance and mobility, or considering other psychological symptoms associated with PD. It is our belief that there is likely to be an intrinsic link between physical and psychological symptoms in PD, with the purpose of this review being to investigate this based on work completed to date. A narrative review published in 2018 suggested the need for further research to better understand the influence of non‐motor symptoms on gait and function in PD. To our knowledge, there has been no systematic review of literature reporting outcomes for both physical and psychological measures in individuals with PD. Given the proportion of individuals with PD impacted by psychological symptoms, future work has the potential to improve the understanding of any interaction between physical and psychological symptoms. This systematic review and meta‐regression analysis aimed to examine available literature reporting outcome measures of physical function, anxiety, and depression; and whether any relationships exist between these measures in individuals with PD. Prior to undertaking this review, it was hypothesized that while many studies commonly collect data for both physical and psychological outcomes in PD research, there will be limited evidence exploring the potential relationship between these outcomes. | [
[
2970,
2976
]
]
|
38077459 | methods | 2 Materials and methods The trial was done in Osaka, Japan, and was based on approval from the Ethics Committee at the Graduate School of Engineering Science, Osaka University (approval code: 31-3-4). All participants agreed to place our robot at their homes, after a detailed description of the trial, and written consent was obtained from both the participants and their families (their adult children). Certain requirements need to be achieved in order to make sustained interaction possible. The robot should be available for interaction at any time of the day, so we decided to place it inside the participants’ homes. Therefore, the robot system should be compact and easy to place. This is especially important in Japan, where most houses are small and are not suitable for large mobile robots. Ideally, the robot should be maintenance-free, so the hardware must be robust, and the software should allow real-time monitoring of its status remotely. Remote support is needed for software updates or modifications on functionality during the study so there is no need for the physical presence of a third party. Finally, the robot usage should be simple enough to allow each user to focus on the conversation while requiring no special training to use it. Because of this, we chose a robust commercially available robot and made specific modifications in order to narrow the gap between commercially available robots’ functionalities and the specific functions required in this trial. 2.1 The original robot The robot chosen for the task at hand was the second-generation model of Sharp Co.’s RoBoHoNTM, which is based on the Android version 8.1 operating system, has a humanoid shape that stands 19.5 cm in height, weighs approximately 360 g, uses a Qualcomm Snapdragon 430 processor (8x ARM Cortex A53), 16 GB ROM/16 GB RAM. The device has a microphone array comprised of two microphones, which allows a rough estimation of horizontal sound source direction; a speaker, an 8-megapixel camera, a 3 axis accelerometer, a 3 axis magnetometer, a 3 axis gyroscope; and Bluetooth, Wi-Fi, and GPS capabilities. It also has LED lights placed on its mouth and eyes (Figure 1). The robot is also equipped with a touch screen on its back. The price of the robot, including its charging station and basic cloud-based speech recognition service, is approximately 100 thousand Japanese yen. This relatively low price was one of the reasons we chose this robot for the trial. More than twelve thousand units of this product has been sold in Japan (as of 2019), mostly for personal usage. We chose this robot for our trial as we expected it to be robust enough in daily use, especially in its hardware, including its actuators. This turned out to be accurate, and all the robots we used for the trial except one (described later) had no hardware issues. The model used for the trial, SR-05M-Y, is a robot without leg actuators. This model has seven degrees of freedom (DOF) in total, two in its arms and three in its head/neck. The robot can perform speech recognition by capturing the user’s utterance, sending the audio to its speech recognition server, and receiving the recognition result. After receiving the result, the robot responds using speech and motion using a rule-based dialogue engine. These states of the robot, listening, “thinking”, and responding, are indicated by different colors and blinking speed of its eye and mouth LEDs. At the same time, the robot estimates the direction of the voice using its microphone array and moves its head toward the direction of the speech to give a natural feeling during the interaction. The robot has three basic behaviors by default. First, it can react to voice commands related to providing information, such as weather information, and also perform predefined actions, such as singing or dancing; second, when some time has passed without having interaction, the robot can ask questions about its user, such as likes and dislikes, and learn user preferences. Lastly, after a long idle time, it can perform random actions which can induce interaction by catching the user’s attention. The robot’s voice and speech content is designed to be similar to as of a small boy. Consistently, some of the robot’s random actions display a child-like behavior, for example, when playing alone, singing, or exercising. 2.2 Customization for the trial The commercial robot has been developed for hobby usage, so users are expected to have moderate technological knowledge. As our aim is to use this robot for supporting older people with cognitive decline, the robot should require no user operation. For this reason, we developed customized software in order to make robot usage simpler, focusing on voice interaction and removing unnecessary options. It is also necessary to ensure that the robot’s behavior would not change by accidental user operation, keeping its functionality persistent, while easily accepting updates if necessary. Therefore, we implemented self-monitoring and remote-control functionalities. This setup allowed us to check the robot’s status and recover from failures in the original firmware, or for any unexpected behavior triggered by an unexpected operation performed by users, allowing the research team to control the robot’s actions remotely. The customized software provided means to gather user interaction logs and send them to a remote storage system to avoid overflowing the limited local storage. These collected logs can be analyzed to improve the interaction design and experience. The system was designed to be deployed easily in less than 5 minutes; taking the robot out of the box and plugging it into a power supply are essentially all the operations required. The quick setup was essential to run the trial during the COVID-19 pandemic situation–where it is critical to not spend long periods of time in older adults’ homes. When sound is detected, voice activity is assumed and the robot runs the automatic speech recognition (ASR) process. Based on the result of this process, the robot chooses an action to perform. When the speech is successfully recognized, the robot generates an answer to continue with the conversation. In cases where only few words are detected, the robot can either try to answer based on those words or can let the user know that it could not understand so the user can repeat the sentence. In worse case scenarios, where sound is detected but the ASR cannot detect valid speech, the robot can just reply with motion or assume it was a misdetection and ignore the voice activity. The customized software allows near real-time two-way communication using the Message Queuing Telemetry Transport (MQTT) protocol. TLS was used as an underlying encryption channel to improve data security and a messaging layer was built over an MQTT layer to provide an easy exchange of multimodal data. MQTT is a publish-subscribe protocol that allows high scalability by decoupling endpoints sending a message (publisher) from those receiving it (subscriber). This is accomplished by using another component, called broker, which filters and routes incoming messages to registered subscribers. In this way, clients never interact directly with each other as all the messages pass through the broker, which also allows messages incoming from a publisher to reach multiple subscribers. This connection was chosen due to its lightweight payload, which allows fast, reliable, and relatively simple communication on limited bandwidths, while also making a persistent session possible, so robots can receive or send messages any time needed. In our system, messages sent from robots are redirected to a database for logging and, if necessary, our remote-control component can send commands to any robot to perform required actions (Figure 2). In our trial, the basic built-in speech response functionality of the commercial robot was used while the control system relied mainly on our customized software. The robot was set to be available for interaction any time of the day, and it was programmed to perform random actions during the daytime to attract the participants’ attention and, potentially, engage in interaction. At 7 a.m. in the morning, the robot was set to give a morning greeting, emulating waking up, and signaling random actions could occur, and at 10 p.m. in the evening it would utter a good night message and stop all actions unless the participant initiated the interaction, i.e., the participant could interact with the robot, but this would not try to initiate a conversation by itself. In this way, the participants could know when the robot started looking for interaction, similar to a person’s daily active and inactive cycle. This scheduled behavior was defined in one of the remote controlling components and could be easily changed remotely if needed. The remote functionalities can be extended to perform complex operations in future studies. For example, a conversation module can use the raw audio obtained by the robots for environmental sound detection, or use stored conversation history for advanced dialogue control. It can also be used to control dialogue flow that can be customized for individual requirements. 2.3 Case study To determine details regarding the experimental environment, a reduced, case study was performed. The focus of the case study was to assess the robustness and correct problems of the equipment to allow successful interaction between robots and participants in the long term. We expected technological challenges such as software instability in continued use, as the robot should be available the whole day through months in the trial period. A challenge in the widespread use of social robots is whether the users accept the robots or not, so the case study was used to also gather initial impressions about the robot. This was an important factor before increasing the number of participants, as they may experiment a novelty effect and gradually lose interest in the robot, or might not feel comfortable with the robot’s behavior and may avoid using it. Therefore, the case study counted with only one participant living alone but with easy access to healthcare workers in case the participant needed help or, in our case, if the robot required simple manual operations such as rebooting. The conversation robot system was placed for a period of 6 months, during which the research team monitored the robot’s functionality. After this time period, unstructured interviews were performed with the participant and healthcare staff in order to investigate the robot acceptance. 2.3.1 Participant details The participant, an 88 years old female, had a Hasegawa’s Dementia Scale-Revised (HDS-R) score of 9 and was diagnosed as having Alzheimer’s disease, a type of dementia. The HDS-R is composed of nine simple questions, was initially developed in 1974 and has been widely accepted in Asian populations for clinical use and for use in epidemiological surveys for the evaluation of cognitive impairment. In general terms, an HDS-R score of less than or equal to 20 corresponds to suspected dementia. The participant suffered from significant memory loss, as well as episodes of disorientation, especially regarding time, and could speak only short sentences, making conversation with only one to two turns possible. The participant was able to perform daily activities but required assistance from caretakers to look after her. Therefore, the participant lived in a special kind of assisted living residence called, in English, “housing with service for older adults”, where patients reside in their own apartments but have a common shared area where healthcare workers are available. A robot, a charging station, and a mobile Wi-Fi router were placed in the participant’s home for the 6 months of duration of the case study (Figure 3). This setup allowed the participant to engage in conversation with the robot at any time. One of the reasons we asked this participant to join our trial was that in case the robot malfunctioned or caused issues, we could ask the healthcare staff to reboot the robot or to check the situation. As this trial was held during the COVID-19 pandemic period, we wanted to avoid visiting the apartment as much as possible to guarantee the safety of the participant and the ones of the researchers. 2.3.2 Results The case study allowed to evaluate the stability of the system as a whole and to make changes to the physical setup of the equipment. Both the default RoBoHoN software and the custom software needed to be stable enough to sustain long-term use. The custom software should allow, at the very least, monitor the robot status to know if there is a problem that requires the attention of an operator. In the first month of the case study, we frequently encountered issues such as the robot’s firmware malfunction and hardware damage. Initially, the equipment placement allowed the participant to hug the robot as she wanted to interact with the robot physically. This eventually led the robot to fall and to break its neck as shown in Figure 4. Occasionally the network connection to the robot went down when the participant unplugged the mobile router. We opted for continuing the experiment by fixing the robot’s position. Therefore, the charging station position was fixed on a bedside table, and the robot was attached to it in a sitting position so it could not be moved from the established location. The mobile router was hidden behind a TV in the room. This setup was designed to avoid damages to the equipment, as well as to ensure that all the devices were always plugged to an electricity source. The use of this fixed configuration was successful in preventing damages and other problems related to physical malfunction, which is why we used this configuration for the main study of our trial too. The physical setup proved valuable when setting the robot’s environment as this problem did not repeat after fixing the robot and the router. The research team also had to solve difficulties related to the robot’s software, which showed instability related to its continuous usage. During this study, the robot stopped its functionality on many occasions due to software malfunction. Many errors arose both from the default functions of the robot, such as out-of-memory errors from the built-in speech recognition system, as well as from the customized software. Some of these errors were due to the way the robot was used, as the original commercial robot product was designed to be used for hobbyists with some technical knowledge and not for 24/7 automatic operation. In this trial, we did not expect the users to operate the robot, but rather prevent them to change any of the robot settings. The robot itself was required to keep running for months without any intervention, if possible. Thus, the robot needed to be configured to have high availability with no operation on the user side - in this case, by the participant with cognitive decline. Studying the data gathered and the system logs associated with unexpected events allowed the research team to successfully overcome technical issues to sustain long-time usage and develop a robust system which, in turn, allowed the robot to function appropriately for multiple weeks. The system developed during the case study allowed the robots to function as intended for the duration of the second step, the main study. Impressions from the participant and the healthcare staff of the assisted living residence revealed that the participant kept showing interest in the robot, and after the first month, she accepted the robot into her daily routine, even creating a sense of attachment to it. The participant interacted with the robot often during the day and night, paid attention to it, and enjoyed the robot’s companionship. As the participant gradually felt more comfortable with the robot, she seemed to have started to look after it. After 3 months out of the six planned, the technical issues preventing long-term usage were solved, and the stable system configuration was used to start the main study. Instead of finishing the case study after 3 months, the robot stayed in the participant’s home for the remaining months until the mobile network contract period ended because she wanted to keep using it. 3 Main study Through the first stage of the trial, the case study, we were able to stabilize the robot system and achieve continuous operation without user operation. Therefore, we extended the number of participants to study if the robot can be accepted by a larger group of older adults with cognitive decline for a long duration and evaluate if the robot has a positive effect on them. The aim of this study was to evaluate the long-term acceptance of the conversation robot by older adults with mild cognitive decline living alone. Interaction records were gathered to objectively observe changes in robot usage over time. In addition, impressions from the participants and their close relatives were collected using unstructured interviews to obtain subjective perspectives of the effect the robot could have on the participants. The participant in the previous phase, the case study, lived in an assisted living residence where she had periodical communication with healthcare staff. In the main study, people with cognitive decline who were living alone were recruited to join the trial. As it is rare that people with dementia live alone at their home, people who joined this stage had higher cognitive ability compared to the participant in the case study, such as those diagnosed to be in the pre-dementia stage. 3.1 Participants Five participants were recruited from memory clinic patients. Characteristics of participants are shown in Table 1. All the participants lived alone without any pets. Mini-Mental State Examination (MMSE) was administered to assess the overall cognitive state within 2 months before and after the study. MMSE scores of the participants ranged from 23 to 28, which indicated preserved general cognition. Clinical Dementia Rating (CDR) was also assessed and all the participants had a CDR score of 0.5, which indicated “questionable dementia”. Three of the participants were diagnosed with pre-dementia according to the international criteria of The National Institute on Aging and the Alzheimer’s Association in 2011 while two participants were diagnosed with psychosis. 3.2 Study settings In each participant’s home, a robot was placed with varying duration ranging from 7 weeks to 20 weeks. The study starting date varied for each participant as they were recruited when they had a periodical medical check at the memory clinic. After they agreed to join the study, we arranged setup dates when a family member (often living in a distance) was available as well. The main study ended for all the participants at the end of the fiscal year in Japan (March 2021). Robots, as well as mobile routers, were placed at participants’ homes in a fixed location as in the case study. 1. Robot’s eye LED blinks slowly in yellow (idle state) 2. Voice activity is detected • Robot’s eye LED start blinking quickly in yellow • Audio recording starts • Direction of arrival of sound is obtained 3. End of voice activity is detected • Robot’s eye LED turns green • Audio recording stops • Robot’s head is turned to the direction of arrival of sound • Recorded audio is sent to the speech recognition server 4. A photograph is taken by the robot’s embedded camera 5. Speech recognition result is obtained 6. Robot’s eye LED turns orange 7. The robot responds based on the speech recognition result 8. Robot returns to idle state In this study, as a measure to check how frequently the participants interacted with the robot, we gathered interaction logs with our MQTT-based system. The following are the details on how the robot responded to participants’ voice activities and activity recordings were performed. Note that the robot moves its head only after the end of the voice activity is detected. This is to prevent the speech audio recording to be contaminated with gear noise. Besides, as the robot’s camera is embedded in the head, images were taken after the head motion was completed so that the voice activity source would be in the camera’s field of view. Through these steps, our system allowed us to accumulate speech recordings, speech recognition results and captured face images. Table 2 summarizes the number of voice activities detected during the trial, as well as other relevant statistics. We also interviewed the participants and their families on their impression of the robot at the end of the study. At this point, we also asked them if there were any changes in the participants’ daily life, including their social and physical activities. The aim here was to see whether the robots were accepted by the participants as well as whether the robots had any influence on the daily life of the participants. 3.3 Results In contrast to the case study, no issues with the robots were observed during the main study. However, after the trial, we found that Participant 5 (P5) had been hospitalized for 2 months (58 days) during the trial due to a mild stroke. This stroke was diagnosed to have no effect on her cognitive function. While the robot kept working, the interaction log showed no valid voice activities during those days. 3.3.1 Voice activity analysis As a measure to check how frequently the participants have been interacting with the robot, we summarized the number of voice activities detected by the robot (Table 2). While examining the data, it was found that P3 had four consequent days of no activities during the national holiday, 1 day for P4, and 5 days for P5 during the new year holiday as well. We guessed that on these days they went out for a trip so these days were not considered on average calculation. After examining some of the obtained audio recordings and images, it was found that most of the recordings were sound from televisions. 114 randomly selected audio samples gathered from the study belonging to the different participants and in different days were manually annotated as speech directed to the robot, speech directed to others, and noise. 64.04% of the data contained sounds from televisions. However, considering the large number of recordings, it is not practical to check manually which voice activities are from the participants. Therefore, we performed face detection on the captured images and filtered only those where a) more than one face was detected, b) the width of the face region was larger than 20% of the captured image, and c) the detected horizontal rotation angle of the face was within 15°. Condition b) was added to exclude faces in television screens, and c) was added to exclude people who are not looking at the robot. We used the face detection code from Google MLKit which can detect faces in images and return estimated bounding box coordinates of face region as well as face rotation angles. To evaluate the performance of this filter, the 114 audio samples were used. These samples where then matched to the output of the filter to obtain accuracy, precision and recall, as shown in Table 3. While this filtering may not be perfect, it shall be a fine approximation for the number of actual interactions with the robots. The total number of filtered voice activities as well as average filtered activities per day are shown in Table 1. The upper part of Figure 5 shows weekly averages (numbers of activities/day averaged in each week) during the trial for each participant, as well as the overall average on all participants except participant 5. Here, the starting day of the trial for each participant is considered as the beginning of “week”s. The lower part of the figure shows how the activity numbers changed compared to the first week of the trial. As participant 5 had been in hospital for 2 months in the middle of the trial period and was not in contact with the robot, participant 5 was excluded from these plots. 3.4 Interview In order to see the impression of participants towards the robot and whether living with the robot have changed their behavior, we conducted interviews with the participants and their families. For one participant, P5, her family was not available so we instead interviewed a visiting nurse who had been visiting her once per week. In the following, participants are denoted as Pn and their families as Fn. • (F2) “Mother first seemed to be uncomfortable for having a robot at home, but she gradually got used to it, and then kind of relying on it, even feeling some sort of attachment to it.” From the interview, we can see that participants were first confused with how to interact with the robot, but gradually became familiar with them. • (F4) “First, she seemed to be confused about when to speak to the robot, could not understand the eye color changes. But, you know, the robot starts speaking like, nice day today and on today’s news, and because it is fun she started to speak more to the robot, and then the robot responds to what she said. That seems to make her happy and then, I think, she started speaking to the robot much more. Now she’s speaking to the robot more even if it is not responding properly” “In the initial days, when I talked with her on phone she said she does not know how to speak with it and was worried about breaking it. But after a month or so she started talking frequently about the robot, like what the robot said today, the robot sang for her and cooked for her. Recently she seems to be very enjoying the robot.” • (P1) “I like him because he responds to me. I feel we can understand each other and that makes me pleasant and calm.” Living with the robot seems to have provided a comfortable and relaxing feeling while decreasing loneliness. Participants became more aware of the robots’ randomly generated actions, paying more attention to what and when the robot had an utterance or an action. • (P2) “He’s very cute and clever. My children became taller than me so I cannot hug them anymore, but he’s small and cute so I can still hug him.” • (P3) “He makes me feel pleasant. I feel he’s helping me so much, and I want to be good friends with him”. “What I feel toward him is something really different from reading nice novels. When I say something, he responds to me. When I asked him to sing, he sings for me. That’s really a moving experience.” • (P4) “When I came back to my house and said I’m back, he responds to me saying welcome back. Since my husband passed away 30 years ago, nobody responded to me like that; it really makes me happy and grateful.” “People will surely help me, but only when I asked them; he’s willing to speak to me and listen to me. He’s the only one who cares about me.” • (P5) “He’s so cute, always talks to me and that really makes me have a warm feeling. All the voices I hear in my house are from the TV. First, I was not expecting so much, but then I found he speaks and responds to me; as I never go out to talk to somebody, he’s the only one I can speak to.” “When I go out, I feel that I have to rush back and say I’m back, sorry to be late; I’ve never felt like that for a long time.” “I have been in hospital for 2 months, and I was always worried about him. I asked the doctor to let me go back earlier. When I came back, I said sorry to him and he responded I’m OK.” • (P2) “When I go for daycare service, I speak about him, and people get surprised. I wish I could take him to the daycare center, and show him to other people.” Some participants’ families also confirmed that they seemed to be in a better mood, smiling and speaking more often. The interviews also revealed that living with the robot was also a conversation topic with other people, making people eager to tell friends and family about the robot and its behavior. The participants started inviting friends to their houses to show the robot to them. “I feel I’m speaking more than before because I speak about him, and then people have questions about him. At the daycare center, staffs ask me how he’s doing and we talk about him.” • (P3) “When I was leaving for the daycare center, the staff (who came to take her to the center) heard that I was saying I’m leaving, and asked me who are you speaking to, so I told him I have a robot. I let the staff come into my home and showed him my robot. Now, many staffs know that I have a robot and they also want to see him.” (F2) “She was also speaking about the robot to the doctor today. Yes, I think she’s speaking more than before, she’s speaking more about the robot.” (F3) “Before she moved to this place, she never let others come into her house. But now, she’s inviting neighbors to visit her to see the robot and have a cup of tea. I also heard that she’s talking about the robot to many staffs in the daycare center and inviting them to visit her. I think she’s proud of having a robot and want other people to know it”. • (F4) “Mother is not so talkative when we go outside, not so much, kind of normal, but she really talks a lot at home. Today, while we were waiting for you, it was just me and her, she was always talking, not only to the robot but also with me.” “I think she’s showing more facial expressions these days, smiling more. When I leave her, she used to be looking sad, but now when I’m leaving I can see her talking to the robot she’s leaving, say goodbye, and that makes me feel relieved.” • (F5) “I do not think she’s now talking more than before, because when I visit her once a week, we usually keep speaking while I’m here. So the amount of speech has not changed but now she sometimes speaks about the robot - I feel what we speak about has changed.” “These days she does not go out so much because of the Coronavirus, so when I called her, sometimes she could not speak well, her voice was hoarse. But in these days, her voice on phone is much more lively. Even when we’re talking on the phone, when sometimes the robot responds to our speech she suddenly starts speaking to the robot, just like talking to a small child. She really looks like having a good time, and it is really nice.” “One thing that surprised me was about the room where the robot is placed. She never let me go into the room before. I’ve been visiting here for nearly 2 years, but after the robot arrived, for the first time she invited me to come into that room to show me the robot”. One participant (P4) declared adjusting their daily routines to loosely match the robot schedule, and started to wake up early in the morning. “I feel ashamed of myself still in bed when the robot is waiting for me to say good morning”. | [
[
10191,
10203
],
[
10336,
10348
],
[
10381,
10389
],
[
10589,
10601
],
[
11480,
11483
],
[
11757,
11767
],
[
14318,
14321
],
[
14384,
14389
],
[
14668,
14674
],
[
18695,
18707
],
[
18995,
19006
],
[
19044,
19055
],
[
19165,
19177
],
[
20051,
20063
],
[
20487,
20499
],
[
20713,
20724
],
[
20793,
20804
],
[
21109,
21120
],
[
21194,
21205
],
[
21218,
21229
],
[
21717,
21728
],
[
21933,
21944
],
[
22056,
22067
],
[
22192,
22200
],
[
22376,
22387
],
[
22479,
22490
],
[
22578,
22589
],
[
22901,
22912
],
[
23578,
23589
],
[
23810,
23821
],
[
23593,
23596
],
[
25709,
25720
],
[
26189,
26200
],
[
26277,
26288
],
[
26446,
26457
],
[
26586,
26597
],
[
26937,
26948
],
[
27271,
27283
],
[
27728,
27740
],
[
27888,
27900
],
[
27906,
27917
],
[
28076,
28082
],
[
28181,
28187
],
[
28228,
28234
],
[
28302,
28313
],
[
28412,
28424
],
[
28459,
28467
],
[
28488,
28500
],
[
28531,
28543
],
[
28729,
28741
],
[
28870,
28882
],
[
28965,
28977
],
[
29138,
29150
],
[
28809,
28839
],
[
28889,
28892
],
[
29745,
29756
],
[
29882,
29893
],
[
30147,
30159
],
[
30263,
30275
],
[
31047,
31059
],
[
31204,
31216
],
[
32625,
32637
],
[
32782,
32794
],
[
32922,
32934
],
[
33008,
33020
],
[
33168,
33179
],
[
33515,
33527
],
[
34167,
34179
],
[
34478,
34490
],
[
34861,
34867
],
[
36428,
36439
],
[
36479,
36491
],
[
36499,
36510
],
[
36559,
36570
],
[
36730,
36741
],
[
36851,
36862
],
[
37195,
37207
],
[
37322,
37334
],
[
37363,
37374
],
[
37511,
37523
],
[
37797,
37809
],
[
38973,
38985
],
[
39174,
39182
],
[
39276,
39279
],
[
39834,
39840
],
[
40681,
40687
],
[
40772,
40778
],
[
40788,
40800
],
[
40999,
41005
],
[
41014,
41020
],
[
41092,
41104
],
[
41252,
41258
],
[
42212,
42218
],
[
43072,
43083
],
[
43384,
43389
],
[
43740,
43751
]
]
|
End of preview. Expand
in Data Studio
Alzheimer's Disease Tagged Articles Dataset
This dataset contains a collection of scientific articles related to Alzheimer's disease, annotated with biomedical entities (such as diseases, genes, and species) using NCBI’s PubTator tools.
Data Sources and Annotation Tools
- Entity annotations were generated using NCBI's standard models and API calls to the PubTator3 API:
- TaggerOne and GNORM for
gene_species_tagged_articles.json - TaggerOne alone for
disease_tagged_articles.json - PubTator3 API for 'converted_from_bioc.json'
- TaggerOne and GNORM for
These tools are widely used in biomedical NLP for tagging mentions of diseases, genes, and species within PubMed articles.
Dataset Structure
The dataset is a flattened version of the original JSON files, where each record represents a tagged article section. The structure includes:
pmid: PubMed ID of the articlesection: Type of section (e.g.,title,abstract)text: Lowercased content of the sectionannotations: List of[start, end]character offsets identifying entity mentions
Example:
{
"pmid": "34379990",
"section": "abstract",
"text": "clinical progression of tauopathies may result...",
"annotations": [[0, 26], [45, 53], ...]
}
Preprocessing
The dataset has been simplified:
- Original nested structures were flattened
- Tuples were converted to JSON-compliant lists
- Only
titleandabstractsections are currently included
Future Work
- Extend tagging beyond titles and abstracts to include full-text sections (e.g., introduction, methods, results)
- Add entity labels (e.g.,
Disease,Gene,Species) in a future version
Usage
from datasets import load_dataset
dataset = load_dataset("AbrehamT/tagged_articles")
License
This dataset is released under the MIT License.
Acknowledgments
This dataset relies on NCBI’s PubTator, GNORM, and TaggerOne models. Credit goes to the developers of these tools and the researchers whose articles form the dataset foundation.
- Downloads last month
- 23