Machine learning assisted Raman microspectroscopy for objective discrimination of breast cancer cells from normal mammary epithelial cells
Histopathology is generally considered a ‘Gold Standard’ in diagnosis of cancers. But it is invasive and time consuming. Moreover, it may not be easy to perform biopsy in many situations. Therefore cytology-based non-invasive diagnosis is recommended. Cytodiagnosis of cancers is normally done by staining and observing cellular morphology under the optical microscope. However, conventional cytology suffers from low sensitivity and specificity. This makes definitive diagnosis difficult. There is a need to develop alternative technologies for reliable cytodiagnosis. Here we propose a new non-invasive, label-free diagnostic technique using Raman spectroscopy and machine learning. In this study, we demonstrate discrimination of breast cancer cells (MCF-7) from normal human mammary epithelial cells (HMEpC) using traditional multivariate analyses such as PCA, LDA, extract pure molecular information from Raman hyperspectral data using MCR-ALS and demonstrate reliable classification by feed forward neural networks.