UNI2-h
A pathology foundation model pretrained via self-supervised learning on large-scale pathological image data. It serves as a general-purpose visual encoder for extracting tissue morphology features and can be applied as a frozen feature extractor in downstream computational pathology tasks.
Timeline (last 90 days)
The study uses frozen pathology foundation model representations from UNI2-h and CONCH for feature extraction without fine-tuning.
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.
The frozen representations of pathology foundation models contain information for predicting rare joint molecular phenotypes, supporting such tasks without fine-tuning.
CONCH uses vision-language contrastive learning to align pathology images with pathology report text; UNI2-h is based on large-scale self-supervised pretraining on pathology images.
All Facts (5)
The study uses frozen pathology foundation model representations from UNI2-h and CONCH for feature extraction without fine-tuning.
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%UnverifiedThe frozen representations of pathology foundation models contain information for predicting rare joint molecular phenotypes, supporting such tasks without fine-tuning.
50%UnverifiedCONCH uses vision-language contrastive learning to align pathology images with pathology report text; UNI2-h is based on large-scale self-supervised pretraining on pathology images.
50%