ANALYSIS OF THE PERFORMANCE OF AN ARTIFICIAL INTELLIGENCE ALGORITHM FOR THE DETECTION OF RADIOGRAPHIC SACROILIITIS IN AN INDEPENDENT COHORT OF AXSPA PATIENTS INCLUDING BOTH NR-AXSPA AND R-AXSPA
Authors: Fabian Proft et al.
In suspected axSpA conventional radiographs of the sacroiliac joints are usually obtained to diagnose sacroiliitis. Also, for the purpose of clinical studies, documenting radiographic sacroiliitis is needed to classify patients as radiographic axSpA (r-axSpA) according to the modified New York Criteria. However, the reliability of the radiologic assessment is rather low. To improve the reproducibility of detection of sacroiliitis the authors have used an artificial intelligence algorithm. The pre-trained neuronal network was used to analyze X-rays of 277 patients with nr-axSpA or r-axSpA. The automated analysis provided an accurate detection of sacroiliitis with a sensitivity of 82% and a specificity of 81%. The positive predictive value was 0.89 and the negative predictive value 0.7.
This study demonstrates that artificial intelligence based X-ray analysis allows to accurately detect sacroiliitis close to expert performance. The introduction of image analysis by neuronal networks will improve the diagnosis and classification of axSpA in the future and will of course have many more potential applications.