[19a-P08-1~9] 4 JSAP-Optica Joint Symposia 2024 (Poster)

English Presentation

[19a-P08-7] 【No-Show】Development of Four Channel based Stokes-Mueller Polarimetry Integrated with Machine Learning for Characterization and Classification of Ductal Carcinoma Tissue

Spandana K U¹, Sindhoora Kaniyala Melanthota¹, Raghavendra U¹, Sharada Rai¹, K K Mahato¹, 〇Nirmal Mazumder¹ (1.Manipal Academy of Higher Education)

Keywords:

Stokes vector,polarization,machine learning

This study presents a Stokes-Mueller polarimetry based imaging system for tissue characterization, specifically for distinguishing normal from cancer regions in ductal carcinoma samples. The fibrous structures present in the tissue enhances the contrast through the measurement of polarization signal, making it easier to analyze images in detail and higher classification accuracy. A dataset is created using various polarization parameters, and validation accuracy of 95.78% and testing accuracy of 94.81% is achieved using machine learning (ML) based support vector machine image classification model. In this study, the average depolarization value of 0.644 in normal regions of tissue, while a slightly higher value of 0.826 is observed. The tumor regions showed a reduction in retardance value from 2.373 to 0.748 as compare to normal. The results demonstrate the potential of ML integrated to Stokes-Mueller polarimetry in investigating detail microstructural characterization of tissue.