Scholarlink

Automated thoracic anomaly detection in chest radiographs using deep learning with YOLOv8 and YOLOv9 models has been presented in the current study. Extensive use was made of the publicly available VinDr-CXR dataset consisting of 18,000 annotated scans to delineate the respective model performance with regard to precision, recall, and mean average precision. These have shown that YOLOv8 outperformed YOLOv9, reaching a high mAP of 32.3% compared with 28.2%, thus proving better in thoracic abnormalities detection. Moreover, the precision-recall trade-off of YOLOv8 is much better, and thus it is much more suitable for clinical use in real-time applications. This again points out the prospect of deep-learning advanced models in enhancing the diagnostic accuracy of medical imaging studies and their considerable clinical benefits.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated Detection of Thoracic Abnormalities in Chest Radiograph with YOLO-Based Models: A Deep Learning Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Md. Asif Imrul,
  • Md Ibtesam Bin Shahid,
  • Sazedur Rahman,
  • Maria Rafique,
  • K. M. Safin Kamal,
  • Ahmed Wasif Reza

摘要

Automated thoracic anomaly detection in chest radiographs using deep learning with YOLOv8 and YOLOv9 models has been presented in the current study. Extensive use was made of the publicly available VinDr-CXR dataset consisting of 18,000 annotated scans to delineate the respective model performance with regard to precision, recall, and mean average precision. These have shown that YOLOv8 outperformed YOLOv9, reaching a high mAP of 32.3% compared with 28.2%, thus proving better in thoracic abnormalities detection. Moreover, the precision-recall trade-off of YOLOv8 is much better, and thus it is much more suitable for clinical use in real-time applications. This again points out the prospect of deep-learning advanced models in enhancing the diagnostic accuracy of medical imaging studies and their considerable clinical benefits.