Developing deep learning algorithms to analyze Raman spectral data for brain cancer diagnosis

Project Description Raman spectroscopy is an optical molecular diagnostic technique that can be used for label-free in situ non-invasive cancer diagnosis. The traditional methods to analyze Raman spectra are mainly based on the intensities of characteristic peaks that are related to underlying biochemicals. However, due to the high dimensionality, and the complexity of the spectral profiles, it is often difficult and subjective to perform the analysis using the traditional methods and distinguish the spectra for different types of tissues. Using machine learning and deep learning methods to analyze the spectra and classify cancerous tissues can overcome these difficulties. The goal of the project is to use deep learning algorithms such as convolutional neural networks to analyze Raman spectra and distinguish human brain cancers at different grades and normal brain tissues. In the project, we will evaluate different methods to process Raman spectra. For example, we will evaluate different analysis methods such as classification of the spectra with and without pre-processing.