Multi-label Deep Learning for Thoracic Disease Co-occurrence in Chest Radiography
DOI:
https://doi.org/10.30871/jaic.v10i3.12840Keywords:
Chest X-ray, Multi-label Classification, MobileNetV2, Class Imbalance, Youden's J Statistic, Computer-Aided Diagnosis, Transfer Learning, Medical Image AnalysisAbstract
Chest radiography (Chest X-Ray) remains the primary imaging modality for thoracic disease diagnosis, yet remains susceptible to misdiagnosis due to anatomical complexity and overlapping pathologies. This study proposes a multi-label Deep Learning framework based on the MobileNetV2 architecture for simultaneous classification of 14 pulmonary pathologies. To address the extreme class imbalance inherent in medical datasets, a two-stage fine-tuning strategy, ColorJitter augmentation, and class weighting (pos_weight) in the Binary Cross-Entropy Loss function were implemented. Furthermore, probability threshold optimization was performed dynamically for each class using Youden’s J Statistic. Ablation study results indicate that the baseline model achieved Mean AUROC of 0.828, while the proposed method achieved Mean AUROC of 0.822. However, this marginal trade-off was strategically accepted to achieve the primary clinical objective: dramatically improving sensitivity (Recall) on critical minority pathologies, including Cardiomegaly (from 73.6% to 85.1%), Fibrosis (64.6% to 72.5%), and Hernia (71.9% to 75.0%). This framework enables simultaneous multi-label classification of 14 pulmonary pathologies using independent sigmoid activations, which inherently supports the detection of co-occurring conditions without enforcing mutual exclusivity. Consequently, the approach demonstrates enhanced clinical utility as a medical screening instrument by substantially suppressing false negative rates for high-risk pathologies. When deployed as a Computer-Aided Diagnosis (CAD) system with appropriate clinical validation, this framework demonstrates the potential to serve as a secondary screening tool in healthcare settings with limited access to specialist radiologists, particularly for detecting high-risk pathologies such as Cardiomegaly and Fibrosis.
Downloads
References
[1] S. Sajed et al., "The effectiveness of Deep Learning vs. traditional methods for lung disease diagnosis using Chest X-Ray images: A systematic review," Appl. Soft Comput., vol. 147, p. 110817, 2023, doi: 10.1016/j.asoc.2023.110817.
[2] E. Çallı, E. Sogancioglu, B. van Ginneken, K. G. van Leeuwen, and K. Murphy, "Deep Learning for Chest X-Ray analysis: A survey," Med. Image Anal., vol. 72, p. 102125, 2021, doi: 10.1016/j.media.2021.102125.
[3] M. R. Hasan, S. M. A. Ullah, and S. M. R. Islam, "Recent advancement of Deep Learning techniques for pneumonia prediction from Chest X-Ray image," Med. Rep., vol. 7, p. 100106, 2024, doi: 10.1016/j.hmedic.2024.100106.
[4] H. Aljuaid et al., "An experimental comparison of Deep Learning models for pneumonia classification from Chest X-Ray images," Biomed. Signal Process. Control, vol. 112, p. 108742, 2026, doi: 10.1016/j.bspc.2025.108742.
[5] R. Arora et al., "Multi-label classification of Vindr-CXR dataset using transfer-learning approach," Procedia Comput. Sci., vol. 259, pp. 522–531, 2025, doi: 10.1016/j.procs.2025.01.050.
[6] L. A. Gonçalves et al., "DualAttentionNet: A Convolutional Neural Network for thoracic disease classification in Chest X-Rays," Procedia Comput. Sci., vol. 256, pp. 797–804, 2025, doi: 10.1016/j.procs.2025.01.082.
[7] N. Dong, M. Kampffmeyer, H. Su, and E. Xing, "An exploratory study of self-supervised pre-training on partially supervised multi-label classification on Chest X-Ray images," Appl. Soft Comput., vol. 163, p. 111855, 2024, doi: 10.1016/j.asoc.2024.111855.
[8] J. Sun et al., "Multi-label Chest X-Ray image classification based on label co-occurrence," Biomed. Signal Process. Control, vol. 119, p. 109930, 2026, doi: 10.1016/j.bspc.2026.109930.
[9] M. Thapa et al., "An explainable deep-learning-based multi-label image classification for Chest X-Rays," Procedia Comput. Sci., vol. 258, pp. 2425–2434, 2025, doi: 10.1016/j.procs.2025.01.250.
[10] Ş. Öztürk, M. Y. Turalı, and T. Çukur, "HydraViT: Adaptive multi-branch transformer for multi-label disease classification from Chest X-Ray images," Biomed. Signal Process. Control, vol. 100, p. 106959, 2025, doi: 10.1016/j.bspc.2024.106959.
[11] H. Bhatt and M. Shah, "Convolutional Neural Network ensemble model for pneumonia detection using Chest X-Ray images," Healthc. Anal., vol. 3, p. 100176, 2023, doi: 10.1016/j.health.2023.100176.
[12] X. Wang et al., "ChestX-ray8: Hospital-scale Chest X-Ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 2097–2106, doi: 10.1109/CVPR.2017.369.
[13] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, "MobileNetV2: Inverted residuals and linear bottlenecks," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2018, pp. 4510–4520, doi: 10.1109/CVPR.2018.00474.
[14] F. M. J. M. Shamrat et al., "High-precision multiclass classification of lung disease through customized MobileNetV2 from Chest X-Ray images," Comput. Biol. Med., vol. 155, p. 106646, 2023, doi: 10.1016/j.compbiomed.2023.106646.
[15] Y. Jin et al., "Deep-learning-based classification of multi-label Chest X-Ray images," Comput. Biol. Med., vol. 157, p. 106683, 2023, doi: 10.1016/j.compbiomed.2023.106683.
[16] G. Chae, J. Lee, and S. B. Kim, "Contrastive learning with hard negative samples for Chest X-Ray classification," Appl. Soft Comput., vol. 165, p. 112101, 2024, doi: 10.1016/j.asoc.2024.112101.
[17] Q. Xu and W. Duan, "DualAttNet: Synergistic fusion of image-level and fine-grained disease attention for multi-label lesion detection in Chest X-Rays," Comput. Biol. Med., vol. 168, p. 107742, 2024, doi: 10.1016/j.compbiomed.2023.107742.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Syahril Alamsyah Zainuddin, I Gusti Ngurah Lanang Wijayakusuma
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License (Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) ) that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
