Acquisition of full Mueller matrix distribution of cancer tissue with near-infrared polarization-probe polarization-imaging and search for its optimal polarization conditions

Polarization imaging offers a promising noninvasive approach for early cancer detection. In this study, the full Mueller matrix distribution is obtained through polarization imaging of a specimen containing both cancerous and healthy tissues. We then propose a method to determine the optimal illumination and imaging polarization states for discriminating between cancerous and healthy tissues in this specimen from the Mueller matrix distribution. Using a machine learning algorithm (naive Bayes classifier), we identified the optimal polarization condition and found that it corresponds to the <i>S</i>₃ observation of right circularly polarized light illumination. Our results demonstrate that the use of specific incident polarization states, especially circular polarization, can significantly increase the contrast between cancerous and healthy tissues, highlighting the potential of polarization imaging for clinical applications.