Comparison of algorithms for detection of active inflammatory lesions in sacroiliitis

Introduction and aim. Artificial intelligence is increasingly being used in the medicine, particularly in radiological diagnosis of diseases such as an axial spondyloarthritis (axSpA). The aim of this study is to compare the available algorithms designed to detect active sacroiliitis in patients with axSpA. Material and methods. Four algorithms, two semi-automated and two full-automated for the assessment of bone marrow ede ma (BME) on magnetic resonance imaging (MRI) of the sacroiliac joints (SIJs) were included in the study. They were described and compared in terms of specificity, sensitivity, and correlation of BME detection findings between AI and experts. Analysis of the literature. Among all automated algorithms, the one created by Bressem et al. had the highest number of ex aminations analyzed in the study, involving 593 MRIs of SIJs. The sensitivity and specificity, as well as the correlation between the AI’s detection of BME versus manual, were not calculated for each algorithm. Rzecki’s algorithm had the greatest sensitivity and specificity for BME detection reaching 0.95 and 0.96, respectively. In addition, its Speraman’s coefficient of correlation be tween manual and automated measurements was 0.866. Conclusion. Each of described algorithms is certainly useful in assessing BME in the MRI examinations of SIJs.