Evaluation of clinical value of interpretable Transformer-based models in computer-aided diagnosis of early-stage pneumoconiosis
环境与职业医学 · 2026-07 · 43 (7) · 844-851
DOI: 10.11836/jeom25477 ↗Kurz gefasst
Fachartikel in der Zeitschrift 环境与职业医学, 2026-07. Verfasst von Ling He und 6 weiteren Autor:innen.
Abstract (Originalsprache)
BackgroundThe diagnosis of early-stage pneumoconiosis relies heavily on physician experience, leading to substantial subjectivity and interobserver variability. The clinical value and interpretability of the Vision Transformer (ViT) architecture for the early identification of pneumoconiosis require systematic validation.ObjectiveTo systematically evaluate the diagnostic performance of ViT-based deep learning models for identifying early-stage pneumoconiosis on digital radiography (DR) chest radiographs, and to enhance the transparency of model decision-making process using the gradient-weighted class activation mapping (Grad-CAM) algorithm, thereby providing a reliable basis for developing computer-aided diagnosis tools.MethodsThis study retrospectively collected high-kilovolt chest radiographs from 1326 dust-exposed workers. The lung fields were automatically segmented using a U-Net model trained with a hybrid loss function combining Dice loss and cross-entropy loss, together with data augmentation techniques. Subsequently, using the segmented lung field regions of interest (ROIs), four ViT-derived models (ViT, SimpleViT, TwinsSVT, and CrossFormer) were trained, and their diagnostic efficacies for the binary classification (normal vs. abnormal) of early-stage pneumoconiosis were compared. The consensus diagnosis of three senior radiologists served as the reference standard. The dataset was randomly divided into a training set (928 cases) and a validation set (398 cases) at a 7∶3 ratio. All classification models were trained using the Adam optimizer for 3000 epochs with an initial learning rate of 0.001, and a cosine decay strategy was employed to dynamically adjust the learning rate. Model performance was compared using DeLong's test, diagnostic consistency was assessed via the Kappa coefficient, and the optimal model was further evaluated by feature visualization analysis using Grad-CAM.ResultsThe U-Net lung field segmentation model demonstrated excellent performance on the validation set, achieving a Dice coefficient of 0.971 and a mean intersection over union (mIOU) of 0.944. In the classification task, the CrossFormer model performed best, yielding an area under the curve (AUC) of 0.962 (95%CI: 0.945, 0.979), with accuracy, sensitivity, specificity, and F1-score of 92.0%, 92.1%, 91.8%, and 0.916, respectively. Its diagnostic performance was superior to the other evaluated models (DeLong's test, P 0.05). Grad-CAM visualization revealed that the model's attention maps aligned closely with the typical radiological distribution of pneumoconiosis, mainly involving the upper and middle lung zones and perihilar regions, corresponding to the actual distribution of small opacities.ConclusionThe CrossFormer model demonstrates excellent diagnostic performance for the early identification of pneumoconiosis. Combined with Grad-CAM-based interpretability analysis, the model shows clinically plausible decision-making patterns, providing a reliable model selection basis and technical reference for the development of computer-assisted diagnostic tools for pneumoconiosis.
Schlagwörter
Zitieren
APA
He, L., Luo, Y., Lei, J., Yin, Q., Yang, D., Zhang, C., & Wang, C. (2026). Evaluation of clinical value of interpretable Transformer-based models in computer-aided diagnosis of early-stage pneumoconiosis. 环境与职业医学, 43(7), 844-851. https://doi.org/10.11836/jeom25477
Vancouver
He L, Luo Y, Lei J, Yin Q, Yang D, Zhang C, et al. Evaluation of clinical value of interpretable Transformer-based models in computer-aided diagnosis of early-stage pneumoconiosis. 环境与职业医学. 2026;43(7):844-851. doi:10.11836/jeom25477
BibTeX
@article{He2026,
author = {He, Ling and Luo, Yan and Lei, Junyan and Yin, Qi and Yang, Deming and Zhang, Changbiao and Wang, Cunjian},
title = {Evaluation of clinical value of interpretable Transformer-based models in computer-aided diagnosis of early-stage pneumoconiosis},
journal = {环境与职业医学},
year = {2026},
volume = {43},
number = {7},
pages = {844--851},
doi = {10.11836/jeom25477}
}
EUVIMED