UNET MOBILENETV2: MEDICAL IMAGE SEGMENTATION USING DEEP NEURAL NETWORK (DNN)

Abstract

In this paper, the framework of polyp image segmentation is developed using a Deep neural network (DNN). Here Unet Mobile NetV2 is considered to evaluate the performance of the image from the CVC-612 dataset for the segmentation method. The proposed model outperformed earlier results. To compare our results two parameters, normally Dice co-efficient and Intersection over Union (IoU) are considered. The proposed model may be used for accurate computer-aided polyp detection and segmentation during colonoscopy examinations to find out abnormal tissue and thereby decrease the chances of polyps growing into cancer. MobileNetV2 significantly outperforms U-Net and MobileNetV2, two key state-of-the-art deep learning architectures, by achieving high evaluation scores with a dice coefficient of 89.71%, and an IoU of 81.64%.