Conference Paper2026

Efficient Multi-Class Thoracic Disease Detection on VinDr-CXR Using YOLOv12 with Class-Balanced Training and Anatomical TTA

Afzal Hossain Alif, Mojtoba Zaman Mantaka, Aiman Saad Hamid, Saadia Binte Alam

2026 IEEE 2nd International Conference on Quantum Photonics, Artificial Intelligence & Networking (QPAIN)

IEEE, pp. 1-6

  • Good Health and Well-being — Relevance: 45%
DOI: 10.1109/qpain69676.2026.11546258URL

Abstract

Chest X-ray (CXR) imaging remains the primary screening modality for thoracic diseases, yet automated systems frequently struggle with concurrent multi-pathology detection and precise localization of subtle or diminutive lesions. Contemporary methods such as YOLOv11-MFF have sought to address these limitations through elaborate architectural enhancements, including frequency-adaptive gates and multi-scale fusion modules. In this work, we contest the premise that such architectural complexity is a prerequisite for leading performance. We show that a standard state-of-the-art detection backbone, coupled with principled data-centric strategies, renders bespoke domainspecific modules unnecessary. Our framework integrates Inverse-Frequency Weighted Upsampling to counteract long-tail class distributions and Anatomically-Constrained Test-Time Augmentation (CXR-TTA) to exploit the bilateral symmetry inherent to the thoracic cavity. On the VinDr-CXR benchmark, our approach attains a mean Average Precision (mAP@0.5) of 0.442, substantially exceeding the prevailing state-of-the-art without any custom neural network components. These results affirm that disciplined data-centric training strategies can outperform intricate architectural engineering in medical object detection.

Keywords

  • thoracic disease detection
  • bilateral symmetry
  • YOLOv12
  • chest radiography
  • lesion localization

CCDS Authors

Saadia Binte Alam, PhD
Saadia Binte Alam, PhD

Saadia Binte Alam, PhD

  • Director, MIRA & WiSE Wings

Associate Professor

Department of Computer Science and Engineering

Independent University, Bangladesh

Published at

HCII, Scientific Reports, Cognitive Computation, Neural Computing and Applications

References

  1. 1.Mingxing Tan, Ruoming Pang, Quoc V. Le. (2020). EfficientDet: Scalable and Efficient Object Detection. , 10778–10787[10.1109/cvpr42600.2020.01079]
  2. 2.Agrim Gupta, Piotr Dollar, Ross Girshick. (2019). LVIS: A Dataset for Large Vocabulary Instance Segmentation. , 5351–5359[10.1109/cvpr.2019.00550]
  3. 3.Rachna Jain, Preeti Nagrath, Gaurav Kataria, V. Sirish Kaushik, D. Jude Hemanth. (2020). Pneumonia detection in chest X-ray images using convolutional neural networks and transfer learning. Measurement, 165, 108046[10.1016/j.measurement.2020.108046]
  4. 4.Hanan S. Alghamdi, Ghada Amoudi, Salma Elhag, Kawther Saeedi, Jomanah Nasser. (2021). Deep Learning Approaches for Detecting COVID-19 From Chest X-Ray Images: A Survey. IEEE Access, 9, 20235–20254[10.1109/access.2021.3054484]
  5. 5.Mohammed A. A. Al-qaness, Jie Zhu, Dalal AL-Alimi, Abdelghani Dahou, Saeed Hamood Alsamhi, Mohamed Abd Elaziz, Ahmed A. Ewees. (2024). Chest X-ray Images for Lung Disease Detection Using Deep Learning Techniques: A Comprehensive Survey. Archives of Computational Methods in Engineering, 31(6), 3267–3301[10.1007/s11831-024-10081-y]
  6. 6.Xiaosheng Wu, Hang Zhang, Junding Sun, Shuihua Wang, Yudong Zhang. (2024). YOLO-MSRF for lung nodule detection. Biomedical Signal Processing and Control, 94, 106318[10.1016/j.bspc.2024.106318]
  7. 7.Yufei Jin, Huijuan Lu, Wenjie Zhu, Wanli Huo. (2023). Deep learning based classification of multi-label chest X-ray images via dual-weighted metric loss. Computers in Biology and Medicine, 157, 106683[10.1016/j.compbiomed.2023.106683]
  8. 8.Yi Lin, Yufan Chen, Kwang-Ting Cheng, Hao Chen. (2023). Few Shot Medical Image Segmentation with Cross Attention Transformer. Lecture notes in computer science, 233–243[10.1007/978-3-031-43895-0_22]
  9. 9.Pei An, Yucong Duan, Yuliang Huang, Jie Ma, Yanfei Chen, Liheng Wang, You Yang, Qiong Liu. (2023). SP-Det: Leveraging Saliency Prediction for Voxel-Based 3D Object Detection in Sparse Point Cloud. IEEE Transactions on Multimedia, 26, 2795–2808[10.1109/tmm.2023.3304054]
  10. 10.Shengnan Hao, Xinlei Li, Wei Peng, Zhu Fan, Zhanlin Ji, Ivan Ganchev. (2024). YOLO-CXR: A Novel Detection Network for Locating Multiple Small Lesions in Chest X-Ray Images. IEEE Access, 12, 156003–156019[10.1109/access.2024.3482102]
  11. 11.Alexey Tolkachev, Ilyas Sirazitdinov, Maksym Kholiavchenko, Tamerlan Mustafaev, Bulat Ibragimov. (2020). Deep Learning for Diagnosis and Segmentation of Pneumothorax: The Results on the Kaggle Competition and Validation Against Radiologists. IEEE Journal of Biomedical and Health Informatics, 25(5), 1660–1672[10.1109/jbhi.2020.3023476]
  12. 12.Tarun Agrawal, Prakash Choudhary. (2022). ALCNN: Attention based lightweight convolutional neural network for pneumothorax detection in chest X-rays. Biomedical Signal Processing and Control, 79, 104126[10.1016/j.bspc.2022.104126]
  13. 13.B.J. Stevens, L. Skermer, J. Davies. (2021). Radiographers reporting chest X-ray images: Identifying the service enablers and challenges in England, UK. Radiography, 27(4), 1006–1013[10.1016/j.radi.2021.03.006]
  14. 14.WenZe Fan, Xingchen Guo, Lei Teng, YuXuan Wu. (2021). Research on Abnormal Target Detection Method in Chest Radiograph Based on YOLO v5 Algorithm. 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI), 125–128[10.1109/cei52496.2021.9574450]
  15. 15.Li Guan, Ruting Zhang, Yi Zhao. (2025). YOLOv11-MFF: A multi-scale frequency-adaptive fusion network for enhanced CXR anomaly detection. PLoS ONE, 20(10), e0334283[10.1371/journal.pone.0334283]
  16. 16.Ran Liu, Yang Zhao, Yaqiong Liu, Xi Chen, Shanshan Cui, Feifei Wang, Lin Yi. (2020). Automated detection of lesion in computer tomography images based on Cascade R-CNN. , 43[10.1117/12.2579644]
  17. 17.Ha Q. Nguyen, Khanh Lam, Linh T. Le, Hieu H. Pham, Dat Q. Tran, Dung B. Nguyen, Dung D. Le, Chi M. Pham, Hang T. T. Tong, Diep H. Dinh, Cuong D. Do, Luu T. Doan, Cuong N. Nguyen, Binh T. Nguyen, Que V. Nguyen, Au D. Hoang, Hien N. Phan, Anh T. Nguyen, Phuong H. Ho, Dat T. Ngo, Nghia T. Nguyen, Nhan T. Nguyen, Minh Dao, Van Vu. (2022). VinDr-CXR: An open dataset of chest X-rays with radiologist’s annotations. Scientific Data, 9(1), 429[10.1038/s41597-022-01498-w]
← All publications