Training time domain transfer improves AI diagnosis on new pathology images
Training-Time Domain Transfer for Generator-Free Diagnostics on Novel Modalities in Computational Pathology
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Medical images come in many types, but AI tools usually only understand common image kinds because of limited data. The authors found a way to train AI to diagnose new types of images by converting old image data into the new style during training. This avoids slow and risky image changes during diagnosis and makes AI more accurate on new images. They tested this on prostate tissue samples and showed better results than other methods.
What this means in practice
- •For clinical pathology teams: Use training-time domain transfer to diagnose cancers on new imaging methods without extra processing at patient diagnosis.
- •For medical ai developers: Create diagnostic models for emerging pathology imaging modalities by augmenting training data with synthetic images matching the new modality style.
Tested on one dataset.
Authors
Julia A. Mielcarz, Andrea Camilloni, Nita Mulliqi, Jessica A. Frank, Anders Blilie, Emilius A. M. Janssen, Martin Eklund, Kimmo Kartasalo
Abstract
Novel imaging modalities promise improved diagnostics, but AI development is constrained by scarce training data. Inference-time domain transfer can address this by mapping novel modalities to established ones, enabling reuse of existing diagnostic models. However, this introduces hallucination risks, computational overhead, and regulatory complexity. Inspired by sim-to-real transfer, we propose training-time domain transfer, generating synthetic target-domain training data from archived routine samples to enable direct diagnosis on real inputs without inference-time generation. Using unstained prostate histopathology as a proof-of-concept, we compare this approach with inference-time virtual staining and conventional H&E-based cancer detection. Training-time transfer achieved higher concordance with the conventional benchmark than virtual staining for both raw (Cohen's $κ$=0.59 vs. 0.37; Gwet's AC1=0.73 vs. 0.29) and deparaffinized tissue ($κ$=0.66 vs. 0.61; AC1=0.73 vs. 0.64), supporting its potential for emerging diagnostic modalities.