MSF-Net: A Lightweight Multi-Scale Feature Fusion Network for Skin Lesion Segmentation

Abstract

Segmentation of skin lesion images facilitates the early diagnosis of melanoma. However, this remains a challenging task due to the diversity of target scales, irregular segmentation shapes, low contrast, and blurred boundaries of dermatological graphics. This paper proposes a multi-scale feature fusion network (MSF-Net) based on comprehensive attention convolutional neural network (CA-Net). We introduce the spatial attention mechanism in the convolution block through the residual connection to focus on the key regions. Meanwhile, Multi-scale Dilated Convolution Modules (MDC) and Multi-scale Feature Fusion Modules (MFF) are introduced to extract context information across scales and adaptively adjust the receptive field size of the feature map. We conducted many experiments on the public data set ISIC2018 to verify the validity of MSF-Net. The ablation experiment demonstrated the effectiveness of our three modules. The comparison experiment with the existing advanced network confirms that MSF-Net can achieve better segmentation under fewer parameters.

Segmentation and classification of skin lesions using hybrid deep learning method in the Internet of Medical Things

“…In Liu et al., 55 NCR‐NET (an acronym for “Neighborhood Context Refinement Network”) obtains 94.01% accuracy on the ISIC‐2017 dataset. MSFNet, a Lightweight Multi‐Scale Feature Fusion Network, obtains 92.17% accuracy on the ISIC‐2018 dataset in the following column 56 . Finally, Kaur and Ranade 57 propose a CNN‐based skin lesion segmentation method that uses group normalization and a mixed loss function to achieve 91% accuracy on the ISIC‐2017 dataset.…”

Leveraging Derma Net for Advanced Skin Cancer Detection in Smart Healthcare Frameworks

“…Using a deep neural network, Shao et al (2023) [10] assessed skin lesion segmentation, MSF-Net, leveraging CA-Net, particularly addressing challenging segmentation tasks such as lesion images through considerable scale disparity, uneven lesion shapes, background with low contrast and blurry borders. Their experimentation utilized the publicly available ISIC2018 dataset to assess MSF-Net's performance.…”

“…True Positivee+False Positivee+False Negativee+True Negativeee (10) Precision and sensitivity: Precision, sometimes referred to as positive predictive value, estimates how well the model predicts the positive outcomes. The Precision formula can be found in equation 11; Sensitivity gauges how well the model can detect every positive sample; so it is also called as true positive rate or recall.…”

DTONet a Lightweight Model for Melanoma Segmentation

“…The following is a brief overview of the current research status on lightweight models. MSF-Net [24] introduces spatial attention mechanisms through residual connections within convolutional blocks, focusing on key regions. Simultaneously, it incorporates multiscale dilated convolution (MDC) modules and multi-scale feature fusion (MFF) modules to extract contextual information across scales, adaptively adjusting the receptive field size of feature maps.…”