Classification Of Biomedical Image Using Intelligent Computing Techniques

Abstract

Research on signal and image processing is a multi-disciplinary application newlineincluding academia, engineering, industry and medicine. It is progressing day by day. newlineOut of all these applications, the input/desired data is either signal or image or both. newlineHowever, the work considered for this dissertation is medical image analysis for cancer newlinedetection. Recent technical advancements support the healthcare through the modern newlineequipment for diagnosis. In most of the disease diagnosis cases, imaging has a major role newlineto represent physiological parameters of the patients. It needs to develop different newlinetechniques for accurate detection and diagnosis. Simultaneously, it has the essential role newlinefor patients suffering from cancer. newlineIn this digital world, AI technology handles most of the problems effectively that newlinegrows exponentially. The key point of this technology is machine learning techniques newlineand is represented as intelligent technique. The growth on this area paves the way to newlineapply in many of the applications. In this work, the models are developed for the newlineclassification using modified variants of machine learning techniques. These are deep newlinelearning algorithm-based convolutional neural network and ensemble neural network. newlineFurther, the procedure is expanded with application of fuzzy logic. newlineInitially, the most informatic gene data for cancer detection is analysed with newlineimage representation. Further the cancer with tumors and without tumors are considered newlineto analyse the proposed techniques. Therefore, three different cancers are considered newlinename as brain cancer through tumor detection and breast cancer through cyst detection. newlineThe non-cystic type cancer is considered as skin cancer. Each cancer is experimented newlinethrough each proposed model and found the substantial improvement in detection and newlinediagnosis. newlineDeep learning-based models provides efficient results for image data, that is newlinefound from earlier works. Gene data is of 1-dimensional type and is converted to 2- newlinedimensional data for image representation using Convex Hull algorithm a