Back to all papers

Tracking temporal progression of benign bone tumors through X-ray based detection and segmentation.

November 11, 2025pubmed logopapers

Authors

Rhyou SY,Bang C,Cho YJ,Bae H,Ha YJ,Baek SY,Lee Y,Kim C,Moon JE

Affiliations (6)

  • Department of Electrical and Computer Engineering, College of Information and Communication Engineering, Sungkyunkwan University, Suwon, 440-746, South Korea.
  • Cleverus Corp, Seoul, 06771, Republic of Korea.
  • Department of Nursing, Honam University, 100 Honamdaegil, Gwangsangu, Gwangju, 62399, Republic of Korea.
  • Department of Orthopedic Surgery, Chosun University College of Medicine, Gwangju, 61453, Republic of Korea.
  • J INTS BIO Inc, Seoul, Republic of Korea.
  • Department of Nursing, Honam University, 100 Honamdaegil, Gwangsangu, Gwangju, 62399, Republic of Korea. [email protected].

Abstract

X-ray is the most widely used imaging modality for the initial diagnosis of bone tumors due to its accessibility and cost-effectiveness. However, the longitudinal comparison of benign bone tumors, particularly for assessing size and shape progression over time, remains largely manual and subjective. In this study, we propose FusionX-BBTNet, a deep learning-based framework that enables automated detection, segmentation, and time-sequential analysis of BBTs from X-ray images. The framework combines YOLO-based object detection with U-Net segmentation, and utilizes a novel wavelet-enhanced dataset to improve contour accuracy. To enable real-world quantification, an OCR-based module is used to extract the X-ray scale bar and compute the pixel-to-length conversion ratio. With this, tumor size and area are calculated in millimeters, and their changes over time are visualized through centroid-based alignment. The proposed method was validated on a dataset of 466 expert-annotated X-ray images, achieving a mean IoU of 0.9376 and a boundary F1 score of 0.9827. In addition to providing reliable tumor localization and measurement, the system supports clinical decision-making by offering intuitive shape and area comparisons. This approach has the potential to complement expert interpretation and improve diagnostic efficiency, especially in environments with limited radiological expertise.

Topics

Journal Article

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.