🧪 Task Description¶

Participants are required to develop automated pathology report generation models that can produce accurate and clinically relevant textual descriptions from gigapixel whole slide images (WSIs). Given a standardized WSI input, the model should generate a structured pathology report aligned with expert-written references. The generated reports will be evaluated based on medical-domain-specific metrics, expert review, and their potential integration into clinical workflows.

The task emphasizes:

1. Slide-level understanding: Extracting informative features from high-resolution WSIs.

2. Clinically aligned text generation: Producing reports that capture diagnostic findings, terminology, and formatting consistent with real-world pathology practice.

3. Robustness across populations: Ensuring generalizability across diverse ethnic and regional data sources.