The CHIMERA challenge aimed to improve predictions for high-risk non-muscle-invasive bladder cancer (HR-NMIBC) using multimodal datasets. Task BRS focused on classifying BCG Response Subtypes from histopathology and clinicopathological data, while Task Progression modeled time-to-progression using histopathology, structured data, and RNA sequencing. The challenge involved 368 patients and saw 159 submissions, with top models achieving a weighted F1 score of 0.73 for Task BRS and a C-index of 0.68 for Task Progression. Post-challenge analysis indicated that histopathology could partially compensate for missing structured data in Task BRS, and progression models benefited from complementary inputs. The study also identified patient-level prediction difficulties, particularly associated with the T1 substage, and highlighted the need for missingness-aware modeling and multi-institutional validation. AI
IMPACT Establishes a benchmark for AI in bladder cancer prediction, highlighting modality contributions and areas for model improvement.
RANK_REASON The item describes a research challenge and its benchmark results on multimodal datasets for a specific medical condition. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
