Related Experiment Video
Updated: Sep 12, 2025
07:24
Autologous Microfractured and Purified Adipose Tissue for Arthroscopic Management of Osteochondral Lesions of the Talus
Published on: January 23, 2018
10.4K
Deep Learning-Enabled Ultrasound for Advancing Anterior Talofibular Ligament Injuries Classification: A Multicenter
Xiaochen Shi1, Haoyan Zhang2, Yu Yuan3
1Department of Trauma and Orthopedics, Peking University People's Hospital, No.11, Xizhimen South Street, Xicheng District, Beijing 100044, PR China (X.S., H.X.).
Academic Radiology
|August 5, 2025
Summary
A new deep learning (DL) model, ATFLNet, demonstrated superior performance in diagnosing anterior talofibular ligament (ATFL) injuries using ultrasound. AI-aided strategies significantly improved radiologist accuracy and reduced diagnostic variability in clinical settings.
Area of Science:
- Orthopedics
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Ultrasound (US) is the primary imaging modality for assessing anterior talofibular ligament (ATFL) injuries.
- Current diagnostic methods can be limited by inter-observer variability and the complexity of ATFL injury classification.
Purpose of the Study:
- To develop and validate a US-based deep learning (DL) model (ATFLNet) for classifying ATFL injuries.
- To evaluate the potential of artificial intelligence (AI) in improving radiologists' diagnostic performance for ATFL injuries.
Main Methods:
- A DL model (ATFLNet) was trained and validated on a large dataset of US images from patients with ATFL injuries and healthy controls.
- Model performance was compared against senior radiologists on external validation datasets.
- An AI-aided diagnostic strategy was developed and tested in simulated and prospective clinical scenarios.
Main Results:
- ATFLNet achieved high accuracy (AUC ≥0.970) across all injury classes in multiple datasets.
- The DL model consistently outperformed senior radiologists in diagnosing ATFL injuries.
- The ATFLNet-aided strategy significantly improved radiologist accuracy and reduced diagnostic variability, especially for junior radiologists.
Conclusions:
- The developed US-based DL model demonstrates superior performance compared to human experts in evaluating ATFL injuries.
- AI-aided strategies show significant promise for enhancing diagnostic accuracy and consistency in real-world clinical practice for ATFL injury assessment.
