Description
Lung adenocarcinoma - the most common type of lung cancer, demands accurate diagnostic grading to guide effective clinical management. Current gold-standard Hematoxylin and Eosin (H&E) staining provides morphological contrast but lacks biochemical specificity, limiting quantitative analysis of tissue subtypes within the heterogeneous lung cancer microenvironments. We developed DeepLuAd, an AI-powered platform integrating label-free stimulated Raman scattering (SRS) microscopy with semantic-guided deep learning, enabling automated tumor grading, cellular-level morpho-chemical quantification, and unsupervised virtual H&E staining. The platform achieved a mean intersection-over-union (mIOU) of 80.43% across tissue subtypes, and a grading concordance rate of 76.2% with clinical diagnoses (16/21 cases). Moreover, it quantifies lipid-to-protein ratio heterogeneity within tumor and stromal regions, revealing biochemical signatures of disease progression. The modular and scalable DeepLuAd framework may also be broadly applicable to diverse solid tumors for AI-enhanced histopathology.
