Deep-learning artificial intelligence (AI) analyses may have a similar capacity to predict rheumatoid arthritis when reading the magnetic resonance imaging (MRI) scans of patients who potentially have early disease compared with visual identification by radiologists and rheumatologists, according to a press release from EULAR. The new findings were presented by Li et al at the EULAR 2023 Congress and simultaneously published in the Annals of the Rheumatic Diseases. In a new study, researchers first trained the AI model to distinguish patients with or without markers of the disease—including erosion, synovitis, and bone marrow edema. The researchers then asked the AI model to interpret the MRI scans of nearly 2,000 patients with early-onset arthritis or clinically suspect arthralgia; they demonstrated that the model’s ability to correctly identify early disease was comparable to that of radiologists and rheumatologists during visual assessment. The researchers concluded that more timely diagnosis of rheumatoid arthritis through automated MRI analysis may allow for earlier interventions to slow the progression of disease.
June 09, 2023