Dual attention based network for skin lesion classification with auxiliary learning. (April 2022)

• Record Type: Journal Article
• Authors: Wei, Zenghui; Li, Qiang; Song, Hong
• Is Part Of: Biomedical signal processing and control. Volume 74(2022)
• Publication Date: 2022-04
• Subjects: Skin Lesion Classification -- Spatial attention -- Channel attention -- Auxiliary Learning -- Dermoscopy image; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103549
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A dual attention mechanism is proposed which can highlight the meaningful local patterns contained in the skin lesion regions, enhancing the feature representation and the interpretability of the proposed network at the same time. An auxiliary learning mechanism which contains auxiliary supervision and KL divergence based regularization is proposed. The KL regularization can make the two auxiliary supervision branches collaborate with each other during training through mutual knowledge transferring, and guide the network to extract the meaningful local pattern features contained in the skin lesion region in a weakly supervised manner. Besides, it brings in strong regularization which makes our proposed network avoid over-fitting when training on the small training data. The proposed network gained the state-of-the-art performance for skin lesion classification, regardless of binary- or multi-classification. Besides, the robustness and interpretability of the proposed network are strong which can promote its clinical application. Abstract: Skin lesion varies greatly in appearance, and its classification task suffers from large inter-class similarity and intra-class variation, thus the subtle differences of local pattern contained in the skin lesion regions are critical for its classification. In this paper, we propose a dual attention based network for skin lesion classification with auxiliary learning. The dual attention mechanism includes the spatial attention (SA) and the channel attention (CA) modules. The SA module can focus on the skin lesion region feature with reduced irrelevant artifacts features. In the subsequent CA module, it first captures the non-local based global feature of the skin lesion region and then generates the feature channels reweighting vector, which is used to further refine the meaningful local pattern feature contained in the skin lesion region. Therefore, the performance and the interpretability of the proposed network are enhanced at the same time. The proposed auxiliary learning contains two auxiliary supervision branches and KL regularization. The KL regularization makes the two auxiliary supervision branches collaborate with each other during training through mutual knowledge transferring. The introduced strong regularization can guide the dual attention mechanism to focus on the meaningful local pattern features in a weakly supervised manner and make the network avoid overfitting on small training data. Without extra training data, our proposed network can outperform current competition winners on several datasets, regardless of binary- or multi-classification. The proposed network is robust enough and owns strong interpretability which promotes its clinical application.