CHIMERA-agent Challenge¶

Combining HIstology, Medical imaging and molEcular data for medical pRognosis and diAgnosis Agent

Prostate cancer care requires radiologists, pathologists, and oncologists to integrate information across multiple modalities over an extended patient journey: from initial MRI screening through biopsy, treatment, and long-term follow-up. Yet most AI tools remain siloed, single-modality systems that rarely reflect real clinical practice, where data are frequently missing, fragmented, or contradictory.

CHIMERA-agent builds on the CHIMERA challenge at MICCAI 2025 by covering the full clinical continuum and explicitly modeling missingness, longitudinal evidence, and cross-modality conflict. The key innovation is a focus on agent-level reasoning: participants must submit not only predictions but structured reasoning traces explaining how available evidence was interpreted and integrated.

Note for participants: You are expected to design decision policies that orchestrate pretrained, modality-specific foundation models. Training models from scratch is not required. Organizers provide baseline tools for prostate zone segmentation, Gleason grading, and MRI-based cancer detection to lower the entry barrier.

Key facts¶