Deep learning enhanced label-free cervical screening via stimulated Raman cytology.

Cervical cancer screening remains pivotal for early detection and effective disease management, yet conventional cytopathological methods relying on stained cell-smear analysis face critical limitations in diagnostic throughput and sensitivity. We present a stain-free Visual-Aided Diagnosis via Stimulated Raman Cytology (VAD-SRC) platform that enables rapid cervical cell screening through simultaneous chemical and morphological profiling. By capturing intrinsic biomolecular contrast via stimulated Raman scattering (SRS) microscopy, our platform establishes malignancy-associated cellular fingerprints through quantitative analysis. Integrated with a deep convolutional neural network architecture, VAD-SRC achieves superb diagnostic performance (98.5 % accuracy, 100 % sensitivity) on an independent test set for binary classification of benign versus malignant cases. Moreover, its high-resolution segmentation function automates the identification of individual cancer cells within a mixture of five cell types: normal cells, leucocytes, low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL), and squamous cervical cancer (SCC) cells. This advancement offers promising potential for cervical cancer screening and visual assessments within cytopathology workflows, enhancing diagnostic efficiency and precision.