Published Paper Details:

Deep learning and image segmentation techniques are widely used in various sectors nowadays but, these methods play a predominant role in diagnosis of various health conditions. Automated and detailed medical image segmentation model have gained popularity since the advent of deep learning techniques specifically fully convolutional neural networks (FCNN). U-Net is one such fully convolutional neural network (FCNN) based image segmentation model, which has proven its efficiency in medical image segmentation over recent years. In this paper, we present the traditional U-Net model in comparison with a novel U-Net model with an additional dropout layer in each convolutional layer. Here, we compare these models considering the dice average values obtained for two different data inputs containing ultrasound images of a carotid artery in various patients. The proposed U-Net model with dropout layer is observed to have better dice average values in comparison with the traditional U-Net model.