Path Analysis of the Cancer Mueller Matrix: From Physical Decomposition to High-Dimensional Vector Mapping

Subtle differences between cancerous and normal regions in unstained tissue sections limit the performance of conventional diagnostic methods. Mueller matrix polarimetry provides comprehensive information on tissue polarization responses; however, the intrinsic coupling of optical effects in the original matrix elements complicates direct histological interpretation. In this study, a pixel-level polarization dataset from clinical lung cancer and basal cell carcinoma sections is established to systematically compare low-dimensional physical decomposition and high-dimensional combination mapping for cancer-region identification. Conventional decomposition approaches exhibit limited discriminative capability and significant class overlap in complex tissues. To address this limitation, a vectorial metric norm spectrum incorporating multi-order features is developed, enabling enhanced representation of polarization characteristics. The proposed high-dimensional mapping framework achieves accurate differentiation between cancerous and normal regions and demonstrates strong performance in cross-validated evaluations. This work establishes a progressive strategy from physical decomposition to high-dimensional representation for optical-assisted pathological analysis.