Related Experiment Video
Updated: Dec 20, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.3K
Incremental inputs improve the automated detection of implant loosening using machine-learning algorithms
Romil F Shah1,2, Stefano A Bini2, Alejandro M Martinez3
1Department of Orthopaedic Surgery, University of Texas, Austin, Texas, USA.
The Bone & Joint Journal
|June 2, 2020
Summary
Machine learning accurately detects prosthetic loosening from radiographs, improving with advanced algorithms and patient history data. This AI tool aids clinical decisions for hip and knee replacements.
Area of Science:
- Orthopedic surgery
- Artificial intelligence in medicine
- Medical imaging analysis
Background:
- Prosthetic loosening is a common complication after total hip (THA) and total knee arthroplasty (TKA).
- Accurate preoperative diagnosis of prosthetic loosening is crucial for successful revision surgery.
- Radiographs are routinely used to assess implant fixation, but interpretation can be challenging.
Purpose of the Study:
- To evaluate a machine-learning algorithm's ability to diagnose prosthetic loosening using preoperative radiographs.
- To identify factors that can enhance the algorithm's diagnostic performance.
Main Methods:
- A convolutional neural network (CNN) was trained on preoperative radiographs from 697 patients undergoing THA or TKA revision.
- The CNN model incorporated patient historical and comorbidity data.
- The final model was tested on an independent dataset to assess accuracy, sensitivity, and specificity.
Main Results:
- A CNN model using only radiographs achieved 70% accuracy.
- The final model, combining radiographs and patient history, achieved 88.3% accuracy, 70.2% sensitivity, and 95.6% specificity.
- The algorithm performed better in diagnosing loosening for revision THA (90.1% accuracy) than for revision TKA (85.8% accuracy).
Conclusions:
- Machine learning, particularly CNNs, can effectively detect prosthetic loosening from radiographs.
- Algorithm performance is significantly enhanced by using pre-trained models (e.g., DenseNet) and integrating clinical data.
- While not a standalone diagnostic tool, this AI approach serves as a valuable adjunct for clinical decision-making in prosthetic loosening assessment.
