Abstract:Skin lesion segmentation is of great significance for skin lesion analysis and subsequent treatment. It is still a challenging task due to the irregular and fuzzy lesion borders, and diversity of skin lesions. In this paper, we propose Triple-UNet to automatically segment skin lesions. It is an organic combination of three UNet architectures with suitable modules. In order to concatenate the first and second sub-networks more effectively, we design a region of interest enhancement module (ROIE). The ROIE enhances the target object region of the image by using the predicted score map of the first UNet. The features learned by the first UNet and the enhanced image help the second UNet obtain a better score map. Finally, the results are fine-tuned by the third UNet. We evaluate our algorithm on a publicly available dataset of skin lesion segmentation. Experiments show that Triple-UNet outperforms the state-of-the-art on skin lesion segmentation.

What problem does this paper attempt to address?

This paper attempts to solve the challenging problems in skin lesion segmentation. Specifically, the boundaries of skin lesions are usually irregular and blurry, and at the same time, skin lesions themselves are diverse, which makes automatic segmentation very difficult. In addition, the limited number of high - quality labeled images also increases the difficulty of training. ### Main problems include: 1. **Large differences in the size, shape and color of skin lesions**: Some lesions are even occluded by hair, increasing the complexity of segmentation. 2. **The blurry boundary between the lesion and the normal skin**: This makes it difficult to accurately distinguish the lesion area from the background area. 3. **The scarcity of high - quality labeled images**: The limited high - quality labeled data makes model training more difficult. To solve these problems, the author proposes a new architecture - Triple - UNet, aiming to automatically segment skin lesions by combining three U - Net structures and appropriate modules. The following are the key points mentioned in the paper: ### Proposed solutions: - **Triple - UNet architecture**: Composed of three U - Net sub - networks, each sub - network has its specific function. The first sub - network generates a preliminary prediction score map, the second sub - network performs more accurate segmentation on the enhanced image, and the third sub - network fine - tunes the result. - **Region of Interest Enhancement module (ROIE)**: By using the prediction score map of the first U - Net to enhance the region of interest of the input image, thereby improving the segmentation performance of subsequent sub - networks. - **Depthwise separable convolution and channel attention module**: Used to improve the encoder and decoder to improve computational efficiency and reduce model parameters. ### Experimental results: The author evaluated Triple - UNet on the ISIC 2018 skin lesion boundary segmentation data set and demonstrated its superior performance in metrics such as Dice coefficient and mIoU. The experimental results show that Triple - UNet significantly outperforms the existing state - of - the - art methods. ### Summary: The main contribution of this paper is to propose a new Triple - UNet architecture, which effectively solves the challenging problems in skin lesion segmentation by combining multiple U - Net sub - networks and the region of interest enhancement module.

A Region of Interest Focused Triple UNet Architecture for Skin Lesion Segmentation

Attention-Guided Network with Densely Connected Convolution for Skin Lesion Segmentation

Segmentation of Liver Tumor based on Mobilenetv3 U-shape Net

SEACU-Net: Attentive ConvLSTM U-Net with squeeze-and-excitation layer for skin lesion segmentation

A Multi-Stage Framework With Context Information Fusion Structure For Skin Lesion Segmentation

EIU-Net: Enhanced feature extraction and improved skip connections in U-Net for skin lesion segmentation

MSS-UNet: A Multi-Spatial-Shift MLP-based UNet for skin lesion segmentation

Skin lesion region segmentation model based on improved U2Net

USL-Net: Uncertainty Self-Learning Network for Unsupervised Skin Lesion Segmentation

DI-UNet: dual-branch interactive U-Net for skin cancer image segmentation

A Medical Image Segmentation Method Based on Improved UNet 3+ Network

UNet 3+: A Full-Scale Connected UNet for Medical Image Segmentation

Skin lesion segmentation with attention-based SC-Conv U-Net and feature map distortion

3D RoI-aware U-Net for Accurate and Efficient Colorectal Tumor Segmentation

DONet: Dual Objective Networks for Skin Lesion Segmentation

GSCEU-Net: an End-to-End Lightweight Skin Lesion Segmentation Model with Feature Fusion Based on U-Net Enhancements

Skin Disease Segmentation Method Based on Network Feature Aggregation Module and Edge Enhanced Attention Mechanism

UCM-Net: A Lightweight and Efficient Solution for Skin Lesion Segmentation using MLP and CNN

NCRNet: Neighborhood Context Refinement Network for skin lesion segmentation

Region Segmentation of Retina OCT Image Layer Based on MS-UNet.

Enhancing skin lesion segmentation with U-Net++: Design, analysis, and performance evaluation