Perceiving the depth of flat and thin adenomas and the extent of invasion is vital in staging, diagnosis, treatment planning, and surgical precision. Raman spectroscopy and its spatially offset version have demonstrated excellent specificity in tissue classification and tumor margin analysis using label-free methods. In this work, we utilize fiber-optic Raman spectroscopy to simultaneously predict the depth and thickness of thin sub-surface tumors using tissue-like multi-layer phantoms. Silicone phantoms incorporating Hydroxyapatite distribution are used as a Raman scatterer to indicate malignant calcifications. The signal intensity for varied tumor depths and thicknesses is also numerically simulated and corroborated with experiments. The high-wavenumber Raman spectrum is captured using a fiber-optic low-resolution spectrometer with an excitation wavelength of 660 nm. The tumor's depth and thickness range from 0.5 mm to 2 mm in 0.5 mm increments. Partial Least Squares Regression (PLSR) analysis is carried out on the acquired dataset for predicting the tumor depth and thickness with a Root Mean Square Error (RMSE) of 0.268 mm (36.33%) and 0.120 mm (12.89%), respectively.
|