The final chapter includes examples of machine learning applications for medical diagnosis using laser molecular spectroscopy and imaging. The spectral data can be collected through a chemical-based or a “profiling”-based approach.

A biopsy is a medical test involving the extraction of a cellular or tissue sample for examination to determine the disease’s presence or extent.1 The term “optical biopsy” introduced in Chapter 2 is commonly understood to represent the use of some form of optical measurements to non-invasively (or minimally invasively) perform diagnosis in vivo and in real time. This chapter’s material is associated with three groups of biological samples and subsequent analysis methods related to breath optical biopsy, liquid optical biopsy, and tissue optical biopsy. Below, the term “volatile organic compounds” (VOCs) is used in the broader sense of endogenous volatile molecular biomarkers of any chemical origin.

The illustrations are confined to in vivo non-invasive diagnostics using laser molecular spectroscopy and imaging in combinations with machine learning. The examples presented here demonstrate the peculiarities of informative feature selection/extraction, data clusterization, and predictive model construction to distinguish between two or more classes.