Choledochal Cancer Region Detection in Hyperspectral Tissue Images Using U-Net

Cholangiocarcinoma (CC), commonly known as choledochal cancer, is the second most frequent type of liver cancer. It primarily affects the elderly population; however younger people are also being diagnosed with it. Hyperspectral imaging (HS) is a spectroscopy-based technique that consists of many images in a band of neighboring spectra, and the reflection of spectra of all pixels is reconstructed to obtain three-dimensional hypercube data. The goal of this study is to provide a solution for semantically segmenting microscopic HS images of choledochal tissues acquired from 174 patients. Principal Component Analysis was used to preprocess the HS images, followed by min max normalization. The proposed system is based on the UNet architecture. The proposed system achieved an accuracy of 68.09% on the training dataset and 61.95% on testing dataset. We also compared the proposed model with other state-of-the-art models. The findings of this study can be used to develop more efficient HS imaging algorithms.