NON-INVASIVE PANCREAS DISORDER DETECTION BY MODIFIED CONVOLUTIONAL NEURAL NETWORK USING IRIS IMAGE

Abstract

The iris image is the most accurate method of identification for a number of diseases, particularly the detection of pancreatic disorders. According to certain publications, a diagnosis of pancreas can be made by looking at the texture of the iris. In light of it, this work proposes to use iris pictures for MCNN-based pancreas dysfunction identification. Pre-processing, Segmentation, feature extraction, and detection phases are all included in the model. First, the best noise-removed images are analysed during the pre-processing stage using three linear filters like Mean Box, Weighted Average, and Gaussian filter. Second, Segmenting process is done by instance segmentation like Single short instance transformed based instance and detection based instance segmentation. The value of the shape, pattern, and texture characteristics is then extracted from the noise-removed image in the third stage utilizing a local Gabor pattern and three pertinent features. Lastly, MCNN performs the detection procedure utilizing these extracted features. The results of the experiment demonstrated the efficacy of the MCNN-based method for detecting pancreatic disorders.