Progression of structural damage is a major negative outcome in rheumatoid arthritis, and evaluation ideally involves radiographic scoring by two readers – but this approach is challenging in large longitudinal cohorts. The RADAR algorithm has been trained and validated in the BCD cohort – where it successfully analysed 7,560 hand and foot joints – before being tested in the ESPOIR cohort – a 20-year collection of X-rays.1,2 The aim was to demonstrate that an AI algorithm can evaluate large volumes of radiographic data. The first ESPOIR evaluations were performed only on good-quality radiographs; in this subset, binary classification achieved an estimated accuracy of 0.95, sensitivity of 0.93, and specificity of 0.99. On behalf of the group, Alain Saraux explained that current efforts are now focused on preprocessing X-rays with artifacts in order to remove superfluous elements, reduce image variability, and standardise inputs – hopefully improving robustness and generalisability.
Magnetic resonance imaging (MRI) is also of value in patients with suspected rheumatoid arthritis, but visual reading requires expertise and is time-consuming. Dennis Ton presented validation results for ADMIRA – an automatic deep-learning MRI analysis system trained on hand and feet images with the visual RA MRI Scoring system as reference from patients with early inflammatory arthritis, clinically suspect arthralgia and symptom-free subjects. To validate the system, it was tested on MRI images from 180 early inflammatory arthritis patients. The findings showed total inflammation, synovitis, tenosynovitis, and osteitis scores had good/excellent agreement with visual scores. Older age and more physical disabilities were associated with higher ADMIRA scores, similar to visual scores. To gain extra insight, ten outliers were visually inspected and heatmaps indicating the location of ADMIRA assessment reviewed. This confirmed that the outliers were mostly due to overestimation of osteitis and that ADMIRA focuses on the same regions as visual readers. Overall, ADMIRA demonstrates good validity.
Ultrasound can also be labour-intensive and requires trained operators. Attendees at the session received an update on the development and validation of a multivariable model for the strategic integration of musculoskeletal ultrasound during diagnostic work up – an initiative aiming to enhance specificity for inflammatory arthropathies by focusing on disease-specific ultrasound lesions, while minimising the time required for image assessment. To do this, a data-driven classification tree was developed using variables from the RADIAL study algorithm, before building ultrasound into Bayesian multivariable models. The integrated algorithm demonstrated high diagnostic accuracy and exceptional negative predictive value across all investigated RMD, suggesting targeted ultrasound examination is a feasible and highly accurate tool – and could be especially useful to exclude a diagnosis of inflammatory arthritis.
Presenting the work, Antonella Adinolfi said “To our knowledge, this study is the first data-driven implementation of ultrasound within the diagnostic work-up of patients with suspected arthritis. These findings suggest that such an approach can effectively prevent overdiagnosis – and therefore overtreatment – of patients with suspected RMD.”
As an alternative to ultrasound, thermography offers a quick, non-invasive way to capture surface temperature changes associated with underlying joint inflammation, but interpretation of thermal images remains subjective. Presenting the work, York Kiat Tan shared an evaluation of different thermal AI imaging models for classifying joint inflammation in rheumatoid arthritis – tested on 200 dorsal hand images from 100 patients. Across all tested algorithms, CatBoost consistently achieved the highest results, with the best balance of accuracy, sensitivity, specificity, and precision – correctly identifying the most positive cases of wrist joint inflammation while maintaining low false-positive rates. It was noted that dominance of joint temperature and texture-related features suggests that physiological heat distribution patterns and micro-texture variations may serve as early, quantifiable indicators of inflammation. The achieved performance underscores the potential of thermal imaging as a rapid and non-contact alternative for joint inflammation screening. Future work will extend the framework to larger cohorts, integrate tools for clinical interpretability, and assess feasibility for routine use.
Another potential rapid and objective assessment of disease activity in inflammatory arthritis is optical spectral transmission imaging (OST) – a non-invasive technique that can detect inflammatory activity in the hands.3,4 Konstantinos Triantafyllias showed good diagnostic performance of OST in the long-term evaluation of patients with inflammatory arthritis – with the tool correlating significantly with clinical arthritis activity markers. This finding was based on OST measurements in 1,312 wrist and finger joints of 60 patients with clinical signs of inflammatory activity at two individual time intervals. Mean OST scores were significantly higher in the inflammatory arthritis group compared to control, with significant correlations with clinical activity markers and swollen and tender joint counts. Longitudinal changes of OST values correlated with changes in disease activity over the same period. This could represent a valuable time- and resource-saving tool to help in disease activity monitoring alongside clinical examination.
Taken together, these varied technological innovations show promise for time-efficient image interpretation in rheumatology.
Source
Pensec H, et al. Artificial Intelligence-Based Analysis of 20-Year Radiographic Progression in the ESPOIR Rheumatoid Arthritis inception Cohort. Presented at EULAR 2026; OP0344. Ann Rheum Dis 2026; DOI: 10.1136/annrheumdis-2026-eular.B.1188.
Ton D, et al. Automatic Deep learning-based MRI analysis of Inflammatory signs in RA (ADMIRA): a validity study of AI-based MRI interpretation in early arthritis. Presented at EULAR 2026; OP0357. Ann Rheum Dis 2026; DOI: 10.1136/annrheumdis-2026-eular.B.571.
Adinolfi A, et al. Point-of-Care Musculoskeletal Ultrasound in the evaluation of suspected arthritis: a data-driven algorithm from the RADIAL Study. Presented at EULAR 2026; OP0354. Ann Rheum Dis 2026; DOI: 10.1136/annrheumdis-2026-eular.B.2638.
Chin K, et al. Artificial Intelligence-Based Thermographic Detection of Joint Inflammation in Rheumatoid Arthritis. Presented at EULAR 2026; OP0343. Ann Rheum Dis 2026; DOI: 10.1136/annrheumdis-2026-eular.B.365.
Triantafyllias K, et al. Monitoring disease activity in arthritis patients using optical spectral transmission: a longitudinal comparison with clinical parameters. Presented at EULAR 2026; OP0356. Ann Rheum Dis 2026; DOI: 10.1136/annrheumdis-2026-eular.B.2523.
References
1. Gandjbakhch F, et al. Multireader assessment as an alternative to reference assessment to improve the detection of radiographic progression in a large longitudinal cohort of rheumatoid arthritis (ESPOIR). RMD Open 2017;3(1):e00034. DOI: 10.1136/rmdopen-2016-000343.
2. Garmendia A, et al. Etude BCD. Revue du Rhumatisme 2025;A12–21.
3. Triantafyllias K, et al. Diagnostic Value of Optical Spectral Transmission in Rheumatoid Arthritis: Associations with Clinical Characteristics and Comparison with Joint Ultrasonography. J Rheumatol 2020;47(9):1314–22. DOI: 10.3899/jrheum.190650.
4. Triantafyllias K, et al. Optical spectral transmission to monitor disease activity in arthritis patients: longitudinal follow-up comparison with clinical parameters. Rheumatology (Oxford) 2025;64(6):3319–27. DOI: 10.1093/rheumatology/keaf007.
About EULAR
EULAR is the European umbrella organisation representing scientific societies, health professional associations and organisations for people with rheumatic and musculoskeletal diseases (RMDs). EULAR aims to reduce the impact of RMDs on individuals and society, as well as improve RMD treatments, prevention, and rehabilitation. To this end, EULAR fosters excellence in rheumatology education and research, promotes the translation of research advances into daily care, and advocates for the recognition of the needs of those living with RMDs by EU institutions.
Contact
EULAR Communications, communications@eular.org
Notes to Editors