Integration Of Fuzzy Logic And Deep Learning For Medical Image Analysis In Neuroimaging

Abstract

Input from magnetic resonance imaging (MRI) scans is used by contemporary deep learning algorithms to assess changes in brain structure. Over the past several years, a variety of deep learning techniques have gained popularity for their application in the process of feature learning and the enhancement of the responsiveness of systems for learning. We are able to apply a deep learning classifier to identify structural changes in the brain and express the spatial information in this study. This is made possible by integrating the transfer learning model with a fuzzy convolutional neural network (CNN) that performed 3D convolutions at earlier levels. Through the utilisation of ResNet and 2D fuzzy convolutional layers, the model makes use of MCI, which is more often referred to as AD, in order to enhance class separability. To facilitate binary classification in the study, the suggested model is trained and evaluated over a collection of sagittal, coronal, or transverse MRI slices. This is done in order to facilitate the job. An approach known as 5-fold cross-valuation is utilised in this research project in order to investigate the accuracy that was achieved via the use of classification issues during testing and training. In terms of accuracy, the suggested technique achieves a rate of 99%, which is higher than both existing methods and methods that do not use ResNet and 2D Fuzzy convolutional layers. This is according to the data obtained from simulations.