SLP-Net:An efficient lightweight network for segmentation of skin lesions

TL;DR

The paper addresses the need for fast, accurate segmentation of skin lesions with minimal computational cost. It introduces SLP-Net, an ultra-lightweight network built on SNP-type convolutions, featuring SLP, SDS, and SFA modules that provide multi-scale feature extraction without a traditional encoder–decoder. On ISIC2018, SLP-Net achieves high accuracy and Dice with about 0.2M parameters and real-time inference, and shows favorable generalization on PH2 when unpretrained. The work demonstrates that SNP-type networks can deliver competitive segmentation performance with drastically reduced compute, enabling deployment on edge devices for rapid clinician support.

Abstract

Prompt treatment for melanoma is crucial. To assist physicians in identifying lesion areas precisely in a quick manner, we propose a novel skin lesion segmentation technique namely SLP-Net, an ultra-lightweight segmentation network based on the spiking neural P(SNP) systems type mechanism. Most existing convolutional neural networks achieve high segmentation accuracy while neglecting the high hardware cost. SLP-Net, on the contrary, has a very small number of parameters and a high computation speed. We design a lightweight multi-scale feature extractor without the usual encoder-decoder structure. Rather than a decoder, a feature adaptation module is designed to replace it and implement multi-scale information decoding. Experiments at the ISIC2018 challenge demonstrate that the proposed model has the highest Acc and DSC among the state-of-the-art methods, while experiments on the PH2 dataset also demonstrate a favorable generalization ability. Finally, we compare the computational complexity as well as the computational speed of the models in experiments, where SLP-Net has the highest overall superiority

Figures (9)

• Figure 1: Images from the ISIC2018, Three common dataset segmentation challenges are demonstrated, namely, hair interference, scale interference, and lesion boundary blurring irregularities.
• Figure 2: Three neurons with different working mechanisms
• Figure 3: The architecture of the proposed SLP-Net.
• Figure 4: The proposed SDS module. The input and output channels for each layer are in parentheses. $C$ denotes concatenate.
• Figure 5: The proposed SLP module. DW Conv and PW Conv denote depthwise convolution and pointwise convolution, respectively. $f(\cdot)$ denotes activation function. $r$ and $d$ denote the expansion rate and the number of zeros to be filled between two neighboring parameters in the convolution kernel, respectively.