Tumor localization in breast thermography with various tissue compositions by using artificial neural network

Abstract

Identifying and treating the tumor at its early stages has become one of the major challenges faced in the area of breast imaging field since the number of women diagnosed with breast cancer has gradually increase over the years. Breast thermography has distinguished itself as a promising adjunctive imaging modality to the current breast imaging standard for early detection of breast cancer. It provides additional information of underlying physiological changes of the cancerous tissues. However, this particular technique has not yet been accepted for clinical use for it is shown to be highly dependent on a trained operator and also due to the unavailability of a large clinical database for reference and classification. Therefore, this study proposed the development of Artificial Neural Network for tumor localization using thermal data obtained from the previous works. It utilized multiple features extracted from a series of numerical simulations conducted on various tissue composition breast models and were fed into the optimized ANN system of 6-8-1 network architecture with a learning rate of 0.2, an iteration rate of 20000 and a momentum constant value of 0.3. Result obtained shows that this newly developed ANN has a high performance accuracy percentage of 96.33% and 92.89% to both testing and validation data respectively.