DWT based sonoelastography prostate cancer image classification using back propagation neural network

In modern days, Cancer is spreading rapidly which requires a significant attention along with its proper detection and identification, which is even more crucial. Attempt should be made to detect it an early stage so that it may be controlled and sometimes cured. But this requires proper diagnosing methods so that the demerits and pains of being diagnosed are minimized among patients. With respect to these recent day diagnosis that are leaning towards the non-invasiveness associating the Computer-aided technologies, bringing many benefits to this specific area of Medical field, we have tried to propose a methodology which will remove the manual interpretation of detecting the affected regions from an image generated from a modern and new diagnosing modality named Sonoelastography(SE). SE images marks regions in the form of color coded patches depending on the Elasticity scores in a particular Region of Interest (RoI). Here we have tried to analyze such images and classify an image that whether it is malignant or not. We have taken color SE images as our principle input to the proposed system. Discrete Wavelet Transform (DWT) is used to identify the most relevant sections of the image after the preprocessing step(s), where generation of two separate images containing the affected regions and the unaffected regions are carried out. Multilevel thresholding has been used to generate the affected and unaffected images from the original image. The images after thresholding are channeled to red and blue components and are applied to DWT from where the Low Low (LL) component is subjected to obtain features which are used finally to classify the images using Back Propagation Neural Network.