Lesion segmentation using U-Net network

TL;DR

This paper presents a U-Net based method for skin lesion segmentation, emphasizing loss function adjustment and post-processing to handle class imbalance and improve contour accuracy in the ISIC 2018 challenge.

Contribution

It introduces specific training adjustments, including a tailored loss function and post-processing steps, for effective lesion segmentation with U-Net.

Findings

Improved segmentation accuracy on ISIC 2018 dataset

Effective handling of class imbalance in training

Enhanced contour precision through post-processing

Abstract

This paper explains the method used in the segmentation challenge (Task 1) in the International Skin Imaging Collaboration's (ISIC) Skin Lesion Analysis Towards Melanoma Detection challenge held in 2018. We have trained a U-Net network to perform the segmentation. The key elements for the training were first to adjust the loss function to incorporate unbalanced proportion of background and second to perform post-processing operation to adjust the contour of the prediction.