A Detection and Segmentation Architecture for Skin Lesion Segmentation on Dermoscopy Images

TL;DR

This paper introduces a two-stage segmentation architecture with optimized training and ensemble post-processing, achieving state-of-the-art results and winning first place in the ISIC 2018 skin lesion segmentation challenge.

Contribution

The paper presents a novel two-stage segmentation method with optimized training and ensemble techniques, setting new performance benchmarks in skin lesion segmentation.

Findings

Achieved state-of-the-art segmentation performance.

Won first place in the ISIC 2018 challenge.

Validated effectiveness of ensemble post-processing.

Abstract

This report summarises our method and validation results for the ISIC Challenge 2018 - Skin Lesion Analysis Towards Melanoma Detection - Task 1: Lesion Segmentation. We present a two-stage method for lesion segmentation with optimised training method and ensemble post-process. Our method achieves state-of-the-art performance on lesion segmentation and we win the first place in ISIC 2018 task1.