Deep learning-based virtual staining, segmentation, and classification in label-free photoacoustic histology of human specimens

This paper presents a deep learning (DL)-based framework for automated virtual staining, segmentation, and classification in label-free photoacoustic histology (PAH) of human specimens. The framework consists of three components: (1) an explainable contrastive unpaired translation (E-CUT) method for virtual H&E (VHE) staining, (2) a U-net architecture for feature segmentation, and (3) a DL-based stepwise feature fusion method (StepFF) for classification. Applied to human liver cancers, the framework achieved promising results. E-CUT produced VHE images highly similar to real H&E images, preserving morphological aspects. Segmentation successfully identified features like cell area, cell number, and nuclear distance. StepFF achieved 98.00% classification accuracy, exceeding conventional PAH classification (94.80%) and reaching 100% sensitivity based on pathologist evaluation. This framework shows potential as a clinical tool for digital pathology.

Chiho Yoon, Eunwoo Park, Sampa Misra, Jin Young Kim, Jin Woo Baik, Kwang Gi Kim, Chan Kwon Jung, Chulhong Kim