Improving accuracy of CADx system by hybrid PCA and backpropagation

Medical images have recently played a significant role in the diagnosis and detection of various diseases. Medical imaging can provide a means of direct visualization to observe through the human body and notice the small anatomical change and biological processes associated by different biological and physical parameters. To achieve a more accurate and reliable diagnosis, nowadays, varieties of computer aided detection (CAD) and computer-aided diagnosis (CADx) approaches have been established to help interpretation of the medical images. The CAD has become among the many major research subjects in diagnostic radiology and medical imaging. In this work we study the improvement in accuracy of detection of CAD system when combined principal component analysis and feed forward back propagation neural network. This work has investigated the ability to improve the CAD system in order to use in detection abnormality even with low cost diagnosis methods (such as mammogram images or X-ray). The results show that the reduction of correlated details within the training data by using the PCA method can enhance the recognition performance. The performance of the neural network diagnostic to discriminate the normal cases from cancerous cases, evaluated by using recognition analysis show a high accuracy in detection. The proposed approach can be considered as a potential tool for diagnosis breast cancer from x-ray and mammography images and prediction for non-experts and clinicians.