Related Experiment Video
Updated: Nov 9, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.1K
Deep learning for detection of radiographic sacroiliitis: achieving expert-level performance
Keno K Bressem1,2, Janis L Vahldiek3, Lisa Adams1,2
1Department of Radiology, Charité - Universitätsmedizin Berlin, Hindenburgdamm 30, 12203, Berlin, Germany.
Arthritis Research & Therapy
|April 9, 2021
Summary
An artificial neural network accurately detects definite radiographic sacroiliitis, a key sign of axial spondyloarthritis (axSpA). This AI tool shows high performance, aiding in axSpA diagnosis and classification.
Area of Science:
- Radiology
- Artificial Intelligence
- Rheumatology
Background:
- Radiographs of sacroiliac joints are crucial for diagnosing axial spondyloarthritis (axSpA).
- Accurate detection of radiographic sacroiliitis is essential for axSpA classification.
Purpose of the Study:
- To develop and validate an artificial neural network (ANN) for detecting definite radiographic sacroiliitis.
- To assess the ANN's performance in identifying axSpA manifestations.
Main Methods:
- Utilized two independent cohorts of axSpA patient radiographs (1553 and 458 images).
- Trained and validated the ANN on the first cohort; tested on the second independent cohort.
- Evaluated ANN performance using AUC, sensitivity, specificity, Cohen's kappa, and absolute agreement.
Main Results:
- The ANN demonstrated excellent performance with AUCs of 0.97 (validation) and 0.94 (test set).
- Achieved high sensitivity (88-92%) and specificity (81-95%) across datasets.
- Showed strong agreement with human readers (Cohen's kappa 0.72-0.79, absolute agreement 88-90%).
Conclusions:
- Deep artificial neural networks can accurately detect definite radiographic sacroiliitis.
- This AI approach is relevant for the diagnosis and classification of axSpA.
