Publication

Abstract : Deep learning (DL) is occupying a crucial part in the applications related to computer vision that might show various remarkable performances throughout the applications. The major obstacle identified in the DL applications is its complexity in computing. So, there needs an algorithm to reduce the computational complexity in its applications. Many diseases are diagnosed from the images taken in CT and MRI scans. This chapter aimed at the early prediction of Alzheimer’s disease (AD), an algorithm that discriminates the mild cognitive impairment (MCI) and cognitive normal (CN) is casted-off, which shows better results in its analysis. Another survival model is used for the improvement in the accuracy of the prediction, followed by the reduction of ring artifacts in the images. Three consecutive methods are implemented one after the other, which generates a corrected image by means of correlation. In another method for the identification of tuberculosis (TB), a CNN mixed with LSTM is utilized. This shows better results and reduces the computational complexity to the core maximum. This is achieved by retrieving the microscopic images collected from infected persons. CNN is trained in such a way to achieve a prediction accuracy of 99.99%. For estimating the feature of the images, a CNN that is trained earlier is used along with a support vector machine (SVM) classifier and the multilayer perceptron (MLP). For establishing the stable learning framework, a deep CNN is presented along with a cascaded structure that will increase the accuracy. On the whole, this chapter describes some of the methods for reducing the computational complexity of the medical image classifications, thus increasing the chance of disease prediction.