Recognition of brain stroke shape using multiscale morphological image processing

Abstract

ABSTRACT Brain haemorrhage is a form of stroke caused by the inflation in the cerebral artery ensuing confined bleeding in the inherent tissues. The MRI images (T1, T2 and FLAIR (fluid-attenuated inversion recovery)) are extracted from Siemen 3T scanner in which the ordinary and haemorrhage-affected strokes are effectively emphasized in this work. The existing such as CNN (convolutional neural network), residual CNN and MCDNN (mobile-cloud deep neural network) are compared from the perspective of PSNR (peak signal-to-noise ratio), SSIM (structural similarity index measure) and MSE (mean squared error), respectively. The PSNR value of this proposed MMNN (multiple nonotonic neural network) is increased by 0.23%, 0.087% and 0.613% compared to CNN, Residual CNN and MCDNN techniques, respectively. The SSIM value is increased by 0.57%, 0.322% and 0.027% compared to CNN, Residual CNN and MCDNN. MSE value is decreased by 8.93%, 2.1457% and 0.316% compared to CNN, Residual CNN and MCDNN, respectively.