Presentation Information
[8p-P08-2]A Lightweight Deep Learning Pipeline for Label-Free Classification of Prostate Tissue via Two-Photon Imaging
〇(DC)Gagan Raju1, Kausalya Neelavara Makkithaya1, Jackson Rodrigues2, Nirmal Mazumder1, Guan-Yu Zhuo2 (1.Department of Biophysics, Manipal School of Life Sciences, Manipal Academy of Higher Education, Manipal, Karnataka 576104, India, 2.Institute of Biophotonics, National Yang Ming Chiao Tung University, Taipei 11221, Taiwan)
Keywords:
Deep Learning,Tumor Microenvironment,Two Photon Fluorescence
Prostate cancer remains one of the most prevalent malignancies among men worldwide, and early, accurate differentiation between malignant and benign tissue remains a clinical challenge, especially when relying solely on conventional histopathology. In this study, we present a label-free, high-resolution imaging and deep learning pipeline for prostate tissue classification using two-photon microscopy and tissue microarray (TMA) samples. Multiphoton images comprising Second Harmonic Generation (SHG) and Two-Photon Fluorescence (TPF) channels were acquired from prostate cancer tissues, normal prostate tissues, and control cores on TMAs. Following imaging, the TPF channel was subjected to preprocessing and deep feature extraction using the MobileNetV3Small architecture. Embeddings from the penultimate layer were then used as inputs for a Support Vector Machine (SVM) classifier that achieved 91% accuracy. This label-free, high-throughput workflow demonstrates the potential of lightweight deep learning models combined with advanced imaging to support reliable, automated prostate cancer diagnosis and screening.