Optical‐Resolution Photoacoustic Microscopy–Based Virtual Staining: A Wavelet‐Enhanced Contrastive Translation Approach With Structure Preservation

Photoacoustic microscopy (PAM) provides label-free and high-resolution imaging capabilities. However, its optical absorption-based contrast differs fundamentally from hematoxylin and eosin (H&E) staining, hindering integration into standard pathological interpretation workflows and limiting clinical translation and adoption. To address this limitation, we employ a low-cost, radiation-free 532 nm optical-resolution PAM (OR-PAM) system to construct a specimen-level cross-modal dataset for colorectal cancer. Based on this dataset, we propose a virtual H&E staining generation workflow that eliminates pixel-level alignment, enabling conversion of single-wavelength OR-PAM image data into H&E-style images with interpretable tissue structures. Quantitative evaluations demonstrate that our method's virtual sections outperform state-of-the-art unsupervised image-to-image translation models in overall structural fidelity, visual authenticity, and distribution consistency across key assessment metrics. This indicates that virtual staining can act as an efficient visualization layer for preliminary pathological assessment, offering a potential translational pathway for PAM in colorectal cancer clinical imaging.