Segmentation of Eczema Skin Lesions Using U-Net

In this paper, a deep learning technique is used to segment skin lesions of Atopic Eczema. A five-layer model based on U-Net architecture is trained to perform segmentation on a dataset with 84 images of Atopic Eczema. The image dimension is set to 512 by 512 pixels and grayscale, red, green and blue color channels are used for segmentation. Two parameters, accuracy and loss serve as the monitoring parameters during the training of the model. For performance evaluation, parameters including accuracy, sensitivity, specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV) are used. An average validation accuracy of 85.16% is obtained by using green channel.