A Deep Convolutional Neural Network Model for Intelligent Discrimination Between Neurodegenerative Diseases from MR Images

Convolutional neural network (CNN), a variant of artificial neural networks that has been used to control a number of applications related to computer vision including medical image analysis. For Alzheimer's (AD), bipolar disorder (BPD), and Parkinson's Disease (PD), intelligent classification of these neurodegenerative disorders requires accurate diagnosis. The architecture in this study is made to address the overfitting issue and speedy convergence using drop out and weight regularisation. With the use of a number of building blocks, including convolution layers, pooling layers, and fully connected layers, CNN is built in such a way that the system automatically learns spatial hierarchies of feature through backpropagation. The frontal lobe, temporal lobe, lentiform nucleus, insular, thalamus, caudate nucleus, parietal, and occipital regions of the brain are the eight regions of the brain whose volumetric characteristics have been taken into consideration. The first order optimisers Root mean square propagation (rmsprop), ADAM (Adaptive Moment Estimation), and Stochastic gradient descent with momentum (sgdm) were used to carry out the comparison analysis. According to experimental findings, the network was better at diagnosing neurodegenerative disorders using the rmsprop optimizer, with an accuracy rate of 99%.