Early cancer detection drastically improves the chances of cure and therefore methodsare required, which allow early detection and screening in a fast, reliable andinexpensive manner. A prospective method, featuring all these characteristics, isvibrational spectroscopy. In order to take the next step towards the development ofthis technology into a clinical diagnostic tool, classification and imaging methods foran automated diagnosis based on spectral data are required.For this study, Raman spectra, derived from axillary lymph node tissue from breastcancer patients, were used to develop a diagnostic model. For this purpose differentclassification methods were investigated. A support vector machine (SVM) proved tobe the best choice of classification method since it classified 100% of the unseen testset correctly. The resulting diagnostic models were thoroughly tested for theirrobustness to the spectral corruptions that would be expected to occur during routineclinical analysis. It showed that sufficient robustness is provided for a futurediagnostic routine application.SVMs demonstrated to be a powerful classifier for Raman data and due to that theywere also investigated for infrared spectroscopic data. Since it was found that a singleSVM was not capable of reliably predicting breast cancer pathology based on tissuecalcifications measured by infrared micro-spectroscopy a SVM ensemble system wasimplemented. The resulting multi-class SVM ensemble predicted the pathology of theunseen test set with an accuracy of 88.9%, in comparison a single SVM assessed withthe same unseen test set achieved 66.7% accuracy. In addition, the ensemble systemwas extended for analysing complete infrared maps obtained from breast tissuespecimens. The resulting imaging method successfully detected and stagedcalcification in infrared maps. Furthermore, this imaging approach revealed newinsights into the calcification process in malignant development, which was notpreviously well understood.
展开▼
