Presentation Information
[8p-P08-3]Quantitative texture analysis and machine learning aided classification of two photon microscopy images of Oesophageal cancer
〇(D)Kausalya Neelavara Makkithaya1, Guan-Yu Zhuo2, Nirmal Mazumder1 (1.Manipal Univ., 2.NYMC Univ.)
Keywords:
Two photon microscopy,Texture analysis,Machine learning
The diagnosis of esophageal squamous cell carcinoma (SCC) and high-grade dysplasia (HGD) remains challenging due to significant histological overlap, frequently resulting in delayed detection, particularly in high-incidence regions such as Asia. Conventional diagnostic approaches often fail to detect subtle alterations in the extracellular matrix (ECM), as they depend largely on subjective morphological assessment. This study employed two-photon microscopy in conjunction with machine learning algorithms to objectively evaluate esophageal tissue samples, utilizing the gray-level co-occurrence matrix (GLCM) method to extract quantitative texture features and training a support vector machine (SVM) classifier for differentiation between cancer stages. The results indicated that collagen architecture exhibited distinct remodelling patterns depending on the disease progression pathway, and the machine learning models demonstrated high accuracy in distinguishing SCC from HGD. These findings underscore the potential of integrating advanced imaging and computational analysis for early and objective diagnosis of oesophagal malignancies, though validation in larger cohorts is warranted to confirm these outcomes.