PCA and Parellel SVM to Optimize the Diagnostic of Breast Cancer Based on Raman Spectroscopy

Abstract

Breast cancer is one of the most deadly diseases in the world; therefore, rapid automated detection of breast cancer in patients is a relevant step in initiating appropriate treatment. In this paper we present a method that optimizes the response time in the automated detection of breast cancer in which a Raman signal is classified as coming from a healthy tissue biopsy or a damaged tissue biopsy. To carry out the detection, we applied Multivariate Component Analysis (PCA) in conjunction with a Classifier (Vector Support Machine (SVM)) in Parallel and from this methods we obtained high correct detection rates, corroborated when comparing the results of the classifier against previous tissue classifications performed by an expert pathologist. We believe that our approach can be applied to other organs of the body where timely detection and classification of cancer can be difficult and of prognostic relevance, such as stomach and pancreas, among others.