Wavelet Based Adaptive Resonance Theory (ART) Neural Network for the Identification of Abnormalities in Mammograms
Golshah A. Naghdy, Yue Li, Jian Wang · APAMI - HIC 1997: Managing Information for Better Health Outcomes in Australia and the Asia Pacific Region: 11 to 13 August 1997, Asia Pacific Association of Medical Informatics, HISA: Conference Proceedings · 1997
Biological mechanism of the human visual system has motivated many researchers in the computer vision areas from the beginning. There exists cortical neurones which respond to specific frequency and orientation present within the field of view. Gabor Wavelets have been proven to be an appropriate tool to simulate the feature detector simple cells in the visual cortex. The information from these cells will derive the higher level cognition processes. Neural Networks have been increasingly used as high level cognition tools. The combination of low-level wavelet based feature extractors and high-level neural network based cognition have been successfully used in natural texture recognition and classification [1]. Mammography is one of the most widely used methods for the early detection of breast cancers in the developed world. Computer aided detection and classification of the patterns and textures present on mammogram is a powerful tool in mass screening where the human resources are stretched to their limits. This work will present a report on the adaptation of a wavelet based neural network natural texture classifier for mammography pattern classification.