Deeply Supervised Skin Lesions Diagnosis with Stage and Branch Attention
This work addresses the need for efficient and effective skin disease diagnosis tools, particularly for mobile or resource-constrained settings, though it is incremental as it builds on existing lightweight network approaches.
The paper tackles the problem of accurate and unbiased skin lesion diagnosis by developing HierAttn, a lightweight neural network that uses deep supervision with multi-stage and multi-branch attention mechanisms, achieving the best accuracy and AUC among state-of-the-art lightweight networks on datasets like ISIC2019 and PAD-UFES-20.
Accurate and unbiased examinations of skin lesions are critical for the early diagnosis and treatment of skin diseases. Visual features of skin lesions vary significantly because the images are collected from patients with different lesion colours and morphologies by using dissimilar imaging equipment. Recent studies have reported that ensembled convolutional neural networks (CNNs) are practical to classify the images for early diagnosis of skin disorders. However, the practical use of these ensembled CNNs is limited as these networks are heavyweight and inadequate for processing contextual information. Although lightweight networks (e.g., MobileNetV3 and EfficientNet) were developed to achieve parameters reduction for implementing deep neural networks on mobile devices, insufficient depth of feature representation restricts the performance. To address the existing limitations, we develop a new lite and effective neural network, namely HierAttn. The HierAttn applies a novel deep supervision strategy to learn the local and global features by using multi-stage and multi-branch attention mechanisms with only one training loss. The efficacy of HierAttn was evaluated by using the dermoscopy images dataset ISIC2019 and smartphone photos dataset PAD-UFES-20 (PAD2020). The experimental results show that HierAttn achieves the best accuracy and area under the curve (AUC) among the state-of-the-art lightweight networks. The code is available at https://github.com/anthonyweidai/HierAttn.