🏥 PathGenIC: Histopathology Image Report Generation with Multimodal In-Context Learning

PathGenIC is a multimodal in-context learning framework designed for automatic medical report generation from histopathology images. It enhances traditional vision-language models by integrating:

• Nearest Neighbor: Dynamically retrieves similar cases for context.
• Guideline: Follows structured diagnostic rules for consistency.
• Feedback: Learns from mistakes to refine report generation.

We fine-tuned Quilt-Llava-v1.5-7b on HistGen Dataset, and achieves state-of-the-art performance with PathGenIC in BLEU, METEOR, and ROUGE-L scores.

🔗 GitHub Repository | 📄 MIDL 2025 (Under Review)

👉 Download the pretrained model and put it at the model_weight folder.