Some investigations on classification and analysis of mammograms for implementation of computer aided medical decision support system

Abstract

Recent developments in the computerized analysis of medical images have enhanced in various diagnostic tasks and provided valuable information for the healthcare experts. Earlier, the analysis of the medical images was performed by radiologists and analysis by human was limited due to the nonsystematic search pattern of humans, the presence of structure noise in the image and the presentation of complex disease state. This has motivated the integration of huge image data and clinical information to develop the computer-aided diagnosis (CAD) system. This approach is considered as a second opinion in detecting lesions, assessing extent of disease and supporting the diagnostic decisions for improving the healthcare systems. In the recent years the mortality due to breast cancer has increased. In the present work Neural Network and the fuzzy classifiers have been used to implement a Computer-aided diagnostic System to assist the medical practitioners. The acquired mammogram images were preprocessed using Alvarez-Mazorra(AM) shock filter and Region of Interest(ROI) was segmented using Region-based thresholding techniques. The texture and shape features have been extracted to obtain the optimal feature set using Multi Objective Genetic Algorithm (MOGA). Multilayer Backpropagation Neural Network with Ant Colony optimization considering particle swarm optimization (MBPN-ACO-PSO) has been used to classify the tumor. The classification accuracy has been compared with various techniques. It has been seen from the results that the use of optimal feature set improves the accuracy to 99.9%. The malignant tumor is subjected to the identification of the stages of cancer. The tumor size is considered for estimating stages of the cancer. This is accomplished using the fuzzy classifier with tumor size as input. The performance of the developed clinical decision support system has been estimated and found that the classification accuracy and the efficiency are high which proves to be a reliable system for clinical pathology. newline newline