TY - JOUR
T1 - Triple attention learning for classification of 14 thoracic diseases using chest radiography
AU - Wang, Hongyu
AU - Wang, Shanshan
AU - Qin, Zibo
AU - Zhang, Yanning
AU - Li, Ruijiang
AU - Xia, Yong
N1 - Publisher Copyright:
© 2020 Elsevier B.V.
PY - 2021/1
Y1 - 2021/1
N2 - Chest X-ray is the most common radiology examinations for the diagnosis of thoracic diseases. However, due to the complexity of pathological abnormalities and lack of detailed annotation of those abnormalities, computer-aided diagnosis (CAD) of thoracic diseases remains challenging. In this paper, we propose the triple-attention learning (A 3Net) model for this CAD task. This model uses the pre-trained DenseNet-121 as the backbone network for feature extraction, and integrates three attention modules in a unified framework for channel-wise, element-wise, and scale-wise attention learning. Specifically, the channel-wise attention prompts the deep model to emphasize the discriminative channels of feature maps; the element-wise attention enables the deep model to focus on the regions of pathological abnormalities; the scale-wise attention facilitates the deep model to recalibrate the feature maps at different scales. The proposed model has been evaluated on 112,120images in the ChestX-ray14 dataset with the official patient-level data split. Compared to state-of-the-art deep learning models, our model achieves the highest per-class AUC in classifying 13 out of 14 thoracic diseases and the highest average per-class AUC of 0.826 over 14 thoracic diseases.
AB - Chest X-ray is the most common radiology examinations for the diagnosis of thoracic diseases. However, due to the complexity of pathological abnormalities and lack of detailed annotation of those abnormalities, computer-aided diagnosis (CAD) of thoracic diseases remains challenging. In this paper, we propose the triple-attention learning (A 3Net) model for this CAD task. This model uses the pre-trained DenseNet-121 as the backbone network for feature extraction, and integrates three attention modules in a unified framework for channel-wise, element-wise, and scale-wise attention learning. Specifically, the channel-wise attention prompts the deep model to emphasize the discriminative channels of feature maps; the element-wise attention enables the deep model to focus on the regions of pathological abnormalities; the scale-wise attention facilitates the deep model to recalibrate the feature maps at different scales. The proposed model has been evaluated on 112,120images in the ChestX-ray14 dataset with the official patient-level data split. Compared to state-of-the-art deep learning models, our model achieves the highest per-class AUC in classifying 13 out of 14 thoracic diseases and the highest average per-class AUC of 0.826 over 14 thoracic diseases.
KW - Attention mechanism
KW - Chest radiography
KW - Deep learning
KW - Thoracic disease classification
UR - http://www.scopus.com/inward/record.url?scp=85093932017&partnerID=8YFLogxK
U2 - 10.1016/j.media.2020.101846
DO - 10.1016/j.media.2020.101846
M3 - 文章
C2 - 33129145
AN - SCOPUS:85093932017
SN - 1361-8415
VL - 67
JO - Medical Image Analysis
JF - Medical Image Analysis
M1 - 101846
ER -