LSSF-Net: Lightweight Segmentation with Self-Awareness, Spatial Attention, and Focal Modulation

Hamza Farooq,Zuhair Zafar,Ahsan Saadat,Tariq M Khan,Shahzaib Iqbal,Imran Razzak

2024-09-03

Abstract:Accurate segmentation of skin lesions within dermoscopic images plays a crucial role in the timely identification of skin cancer for computer-aided diagnosis on mobile platforms. However, varying shapes of the lesions, lack of defined edges, and the presence of obstructions such as hair strands and marker colors make this challenge more complex. \textcolor{red}Additionally, skin lesions often exhibit subtle variations in texture and color that are difficult to differentiate from surrounding healthy skin, necessitating models that can capture both fine-grained details and broader contextual information. Currently, melanoma segmentation models are commonly based on fully connected networks and U-Nets. However, these models often struggle with capturing the complex and varied characteristics of skin lesions, such as the presence of indistinct boundaries and diverse lesion appearances, which can lead to suboptimal segmentation <a class="link-external link-http" href="http://performance.To" rel="external noopener nofollow">this http URL</a> address these challenges, we propose a novel lightweight network specifically designed for skin lesion segmentation utilizing mobile devices, featuring a minimal number of learnable parameters (only 0.8 million). This network comprises an encoder-decoder architecture that incorporates conformer-based focal modulation attention, self-aware local and global spatial attention, and split channel-shuffle. The efficacy of our model has been evaluated on four well-established benchmark datasets for skin lesion segmentation: ISIC 2016, ISIC 2017, ISIC 2018, and PH2. Empirical findings substantiate its state-of-the-art performance, notably reflected in a high Jaccard index.

Computer Vision and Pattern Recognition,Artificial Intelligence

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

This paper aims to address key challenges in skin lesion segmentation, particularly in computer-aided diagnosis (CAD) on mobile platforms. Specifically, the paper proposes a lightweight network structure (LSSF-Net) for accurately segmenting skin lesion images. The main problems addressed include:

1. **Complex Lesion Shapes**: Skin lesions have variable shapes and blurred edges, making segmentation difficult.
2. **Image Interference Factors**: Elements such as hair and marker colors interfere with the identification of lesion areas.
3. **Subtle Differences**: The texture and color differences between lesions and surrounding healthy skin are subtle and hard to distinguish.
4. **Limitations of Existing Models**: Current models based on fully connected networks and U-Net perform poorly in capturing complex and diverse lesion features.

By introducing self-aware local and global spatial attention, focus modulation attention based on conformers, and channel reordering mechanisms, LSSF-Net can maintain high accuracy while being lightweight, making it suitable for deployment on mobile devices with limited computational resources. The model performs excellently on multiple benchmark datasets (ISIC 2016, ISIC 2017, ISIC 2018, PH2), particularly in terms of the Jaccard index.