State-Aware Interaction MIL
A weakly supervised computational pathology method based on multiple instance learning. It maintains separate representation branches for each biomarker, explicitly models inter-biomarker interactions, and supervises over complete four-state molecular configurations to enable fine-grained prediction of rare joint molecular phenotypes.
Timeline (last 90 days)
State-Aware Interaction MIL is a weakly supervised method that preserves biomarker-specific representations, explicitly models interactions between representations, and performs supervised learning on four-state molecular configurations (double-negative, two single-positives, and joint double-positive).
The study validated the method in two clinical scenarios: joint BRAF+/MSI+ prediction in colorectal cancer and joint EGFR+/TP53+ prediction in lung adenocarcinoma.
Based on UNI2-h representations, State-Aware Interaction MIL achieves an AP of 0.5566 on colorectal cancer, surpassing NaiveMTL's 0.5161.
Based on UNI2-h representations, the method achieves an AP of 0.2784 on lung adenocarcinoma, outperforming IndependentPair's 0.2525.
All Facts (4)
State-Aware Interaction MIL is a weakly supervised method that preserves biomarker-specific representations, explicitly models interactions between representations, and performs supervised learning on four-state molecular configurations (double-negative, two single-positives, and joint double-positive).
50%UnverifiedThe study validated the method in two clinical scenarios: joint BRAF+/MSI+ prediction in colorectal cancer and joint EGFR+/TP53+ prediction in lung adenocarcinoma.
50%UnverifiedBased on UNI2-h representations, State-Aware Interaction MIL achieves an AP of 0.5566 on colorectal cancer, surpassing NaiveMTL's 0.5161.
50%UnverifiedBased on UNI2-h representations, the method achieves an AP of 0.2784 on lung adenocarcinoma, outperforming IndependentPair's 0.2525.
50%