Raman Spectroscopy of Cancer Cells: An Explainable Classification Model

Cancer is a medical illness characterised by the uncontrolled multiplication of cells that can invade the body's normal organs and tissues, changing their form and function. As a result, since cancer is caused by DNA changes within cells, Raman spectroscopy can be a useful method for determining their composition. Using this approach, a sample is illuminated with a monochromatic light beam and the interaction between them causes an effect that allows information on the material under examination to be obtained. The experiments that led to the writing of this work were conducted on a dataset provided by the Center for Nanophotonics and Optoelectronics for Human Health (CNOS); the data concerns the analysis of cells from a patient suffering from liver cancer. The data collection process is very complex and involves several phases, the main one being the analysis of cells using Raman spectroscopy. In the final dataset, 364 frequencies are present. The target variable, which indicates the kind of cell, can have two values: Tumour if the measurement was done on a tumour cell, and Non-Tumor if the cell is healthy. The measurement samples for each record represent the amplitudes corresponding to the different frequencies. The goal of this study is to first build an explainable classification model combined with Raman spectroscopy, and then provide a set of frequencies that best explain the choices the model makes in the decision-making process.