Classification of liver tissue pathological changes via optical biopsy based on refractive index sensing

Area of Science:

• Biomedical Optics
• Medical Diagnostics
• Fiber Optic Sensing

Background:

• Optical biopsy offers minimally invasive tissue assessment but requires rapid, objective analysis.
• Compact sensors are crucial for clinical translation of optical biopsy techniques.

Purpose of the Study:

• To develop a refractive index (RI)-driven classification framework for optical biopsy using fiber-optic interferometry.
• To achieve accurate and explainable diagnostic signatures for liver tissue using machine learning.

Main Methods:

• Utilized an extrinsic fiber-optic Fabry-Pérot interferometric cavity to measure reflection spectra over a biologically relevant RI range (1.33-1.42).
• Defined three proxy classes (healthy, HCC-like, metastatic) based on literature-guided RI windows.
• Engineered 62 spectral descriptors and employed machine learning models (tree ensembles, SVM) and a fuzzy expert system for classification.

Main Results:

• Physics-only RI estimation achieved 0.70 accuracy and 0.48 macro-F1.
• Machine learning models significantly enhanced classification performance, with tree ensembles reaching a macro-F1 of 1.00.
• Feature attribution confirmed RI as the primary discriminative signal, with spectral visibility metrics improving robustness.

Conclusions:

• Machine learning-augmented fiber-optic interferometry provides accurate and explainable diagnostic signatures for optical biopsy.
• The developed framework supports the translational potential of RI-based optical biopsy for liver tissue analysis.