Related Experiment Video
Updated: Jun 4, 2025
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Diagnostic Accuracy and Interobserver Reliability of Rotator Cuff Tear Detection With Ultrasonography Are Improved
Kuan-Ting Wu1, Po-Cheng Chen2, Wen-Yi Chou3
1Department of Orthopedic Surgery, Kaohsiung Chang Gung Memorial Hospital, Graduate Institute of Clinical Medical Science, Chang Gung University College of Medicine, Kaohsiung, Taiwan; Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan, Taiwan; Department of Sports, Health and Leisure and Graduate Institute of Sports, Health and Leisure, Cheng Shiu University, Kaohsiung, Taiwan; Kaohsiung Municipal Fong Shan Hospital - Under the Management of Chang Gung Medical Foundation, Kaohsiung, Taiwan.
This study enhanced YOLOv7 with CBAM (YOLOv7-CBAM) for improved rotator cuff tendon tear detection in ultrasound images. The AI model significantly boosted diagnostic accuracy and consistency for physicians.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Orthopedics
Background:
- Rotator cuff lesions are a common cause of shoulder pain.
- Accurate diagnosis of rotator cuff tears is crucial for effective treatment.
- Ultrasound is a widely used imaging modality for shoulder assessment.
Purpose of the Study:
- To develop and evaluate an improved object detection model, YOLOv7-CBAM, for automated identification of torn or intact rotator cuff tendons.
- To enhance the diagnostic accuracy of ultrasound in detecting rotator cuff lesions.
- To assist physicians in diagnosing rotator cuff injuries more effectively.
Main Methods:
- A modified YOLOv7 model incorporating the convolutional block attention module (CBAM) was developed (YOLOv7-CBAM).
- Ultrasound images from 280 patients with shoulder pain were used for training and testing.
- Performance was evaluated using metrics like accuracy, precision, sensitivity, F1-score, and mean average precision, with gradient-weighted class activation mapping for visualization.
Main Results:
- YOLOv7-CBAM achieved higher accuracy (0.98) compared to YOLOv7 (0.96).
- The model demonstrated superior precision (0.98) and sensitivity (0.98) over YOLOv7.
- AI-assisted diagnosis using YOLOv7-CBAM improved physician accuracy from 80.86% to 88.86% and interobserver reliability from 0.49 to 0.71.
Conclusions:
- The YOLOv7-CBAM model accurately detects rotator cuff tendon tears from ultrasound images.
- Integration of this AI tool can enhance diagnostic accuracy and interobserver reliability for physicians.
- Attentional deep learning models offer a promising approach for improving ultrasound-based rotator cuff diagnosis.
