A Hybrid Approach to Breast Cancer Diagnosis

Abstract

In vivo 31 P Magnetic Resonance Spectroscopy (MRS) is a non-invasive technique for the observation of phosphorus-containing metabolites and intracellular pH. MRS plays an important role in the investigation of cell biochemistry and offers a reliable means for detection of metabolic changes in breast tissue. However, the scarcity of 31 P MRS data and the complexity of interpretation of relevant physiological information impose extra demands that preclude the applicability of most statistical and machine learning techniques developed so far. To overcome such constraints, we propose Knowledge-Based Artificial Neural Networks (KBANNs) [1], a hybrid methodology that combines knowledge from a domain in the form of simple rules with connectionist learning. This combination allows the use of small sets of data (typical of medical diagnosis tasks) to train the network. The initial structure is set from the dependencies of a set of known domain rules and it is only necessary to refine these rules by training. In this chapter, we present KBANNs with a topology derived from knowledge elicited from the domain of metabolic features of normal and malignant mammary tissues. KBANN performance is assessed over the classification of 26 in vivo 31 P spectra of normal and cancerous breast tissues. Results confirm the suitability of KBANNs as a computational aid capable of classifying complex and limited data in a medical domain. The present study is part of an ongoing investigation into normal and abnormal breast physiology, which may help in the non-invasive early detection of breast cancer [2][3].