Background
CHIMERA-Agent aims to establish a multimodal AI benchmark and decision-support framework that advances prostate cancer care by moving beyond single-modality, task-specific models toward an integrated, reasoning-capable agent. By combining histopathology, radiology, and clinical data across the full clinical decision pathway from diagnosis and risk stratification to recurrence prediction and treatment selection CHIMERA-Agent addresses a critical gap: the absence of standardized benchmarks that evaluate not only predictive performance but also the quality, consistency, and interpretability of AI reasoning in a clinically meaningful context.
Aim
Prostate Cancer Recurrence Prediction (Task 1): Current assessments rely on PSA levels, a controversial biomarker with limited reliability in predicting biochemical recurrence. CHIMERA integrates multiparametric MRI and H&E-stained histopathology slides, routinely used in clinical practice, to enhance recurrence prediction. Deep learning has shown that morphological features in histopathology provide significant prognostic value.
Non-Invasive Bladder Cancer (NMIBC) Response Predictions (Task 2 and 3): 50% of NIBC patients fail to respond to BCG treatment, a globally scarce resource. Molecular RNA analysis has been linked to high-risk treatment response and can aid in early treatment decision-making. Integrating transcriptomics, pathology, and imaging can improve prognostic accuracy and patient management.