Sara E. Kochanny

FEATMAP corrects acquisition signatures in medical foundation model embeddings to improve cross-condition similarity and task generalization without retraining.

This paper introduces HistoXGAN, a GAN that reconstructs cancer histology from latent features, showing it retains tumor grade, histologic subtype, and gene expression patterns across 29 cancer types.

This paper shows that digital histopathology batch effects in TCGA data affect DL model accuracy and bias, and proposes a method to address this issue.

This paper develops a deep learning method to predict genetic alterations from histological images across multiple cancer types.