Radiology in Rheumatoid Arthritis
Radiology plays a central role in the diagnosis, monitoring, and prognostication of rheumatoid arthritis (RA), a chronic systemic autoimmune disease affecting approximately 1% of adults in Northern Europe and the United States (roughly 0.24% worldwide), with a female preponderance and typical onset between 45 and 65 years of age.[^c9] The disease is characterized by inflammatory synovitis that, if untreated, leads to progressive joint destruction, pain, and functional impairment. Rapidly evolving treatment regimens, including conventional and biologic disease-modifying antirheumatic drugs, have fundamentally changed the role of imaging by creating a need for earlier and more sensitive detection of inflammation before irreversible structural damage occurs.[^c1]
Conventional radiography remains the most widely available first-line imaging modality for RA, providing a permanent record of joint pathology used for baseline assessment and longitudinal monitoring. However, radiographs are insensitive to early inflammatory changes, with erosions often not visible until one to two years after disease onset. The RAM-H1200 benchmark dataset (1,200 hand radiographs from six centers) now provides standardized annotations including pixel-level bone erosion masks and Sharp/van der Heijde scores, enabling rigorous evaluation of automated radiographic analysis across centers.[^c11]
Ultrasound and MRI have been incorporated into the 2010 ACR/EULAR classification criteria for RA, enabling earlier diagnosis. Ultrasound provides real-time, dynamic assessment of superficial joints and tendons and can distinguish active from inactive synovitis through power Doppler signal. Superb microvascular imaging (SMI), a novel Doppler technique, detects synovial vascular flow in 60.4% of inflamed joints compared to 25.6% with conventional power Doppler.[^c4] Deep learning applied to hand radiographs can predict which finger joints will show ultrasound-detected inflammation, reducing the number of joints requiring ultrasound examination by approximately 80% while improving sensitivity over clinical examination alone.
MRI is the most sensitive modality for detecting bone marrow edema, which is the strongest predictor of future erosive progression. Zero echo time (ZTE) MRI has emerged as a radiation-free alternative to CT for bone erosion detection, achieving 95.9% sensitivity and 97.6% specificity with CT as the reference standard, significantly outperforming conventional MRI and radiography.[^c5] Hybrid PET/MR imaging simultaneously visualizes bone marrow edema and synovitis, combining metabolic and anatomical information in a single examination.[^c10]
Advanced CT techniques have expanded the role of this modality. High-resolution peripheral quantitative CT (HR-pQCT) detects more than 20 times the number of new erosions compared to radiography over one year of follow-up, revealing erosive progression in 40 patients versus only 3 by radiography.[^c7] Dual-energy CT (DECT) shows promise for detecting bone marrow edema as an independent predictor of future erosive progression.[^c8] AI-based CT quantification has identified quantifiable small airway loss and a vascular simplification pattern in RA lungs, with a combination of small airway, small vessel, and interstitial metrics achieving an AUC of 0.837 for detecting RA-associated lung involvement.[^c17][^c18]
Nuclear medicine techniques provide whole-body molecular imaging. FAPI-PET/CT targets fibroblast activation protein on activated fibroblast-like synoviocytes, offering higher signal-to-background ratios and better delineation of individual joints compared to FDG-PET, and can detect subclinical joint inflammation and predict treatment response at six months.[^c13]
Fluorescence optical imaging (FOI) using indocyanine green visualizes impaired microcirculation in both hands in a single six-minute examination, showing greater than 80% agreement with MRI and ultrasound.[^c3] A 2025 study developed a 26-feature FOI atlas enabling a two-step diagnostic pathway that distinguishes RA from osteoarthritis, psoriatic arthritis, and connective tissue diseases. Differential diagnosis on hand radiographs must also account for metabolic deposition mimics: hemochromatosis arthropathy, which preferentially involves the second and third metacarpophalangeal joints with hook-like metacarpal osteophytes, can be difficult to differentiate from RA, though osteophytosis is rarely seen in RA.[^c20] In the crystal arthropathies, the 2023 EULAR recommendations endorse ultrasound and dual-energy CT for gout diagnosis, with characteristic features such as the double-contour sign or tophi on imaging obviating the need for synovial fluid confirmation.[^c21]
Photoacoustic tomography (PAT) is an emerging non-ionizing technology that uses laser-generated ultrasound waves to visualize microvasculature up to 15 mm deep. A 2024 study from University College London developed a hand-held 3D PAT scanner 100 to 1,000 times faster than previous devices, capable of imaging all 20 finger joints in a few minutes. The technology is projected for clinical deployment within three to five years subject to further testing.[^c12]
Artificial intelligence applications are transforming quantitative image analysis in RA. The autoscoRA deep learning system, trained on 12,144 radiographs from 769 patients, achieved an intraclass correlation coefficient of 0.9 for the total Sharp/van der Heijde score against an expert human reader. In a head-to-head comparison, autoscoRA (ICC 0.94) outperformed a second human reader (ICC 0.86) in agreement with the primary reader.[^c6] AI-based quantification of bilateral joint space asymmetry on hand radiographs predicts joint space narrowing progression with AUCs up to 0.836. A dedicated overview of artificial intelligence in RA imaging covers deep learning detection frameworks, automated scoring, and the barriers to clinical adoption.
Imaging has also advanced understanding of extra-articular involvement. Cervical spine abnormalities are present in 75% of RA patients, with anterior atlantoaxial subluxation (58%) and subaxial subluxation (58%) being the most common findings.[^c14] Ultrasound of the dorsalis pedis artery has demonstrated a ≥96% prevalence of wall thickening in RA patients, far exceeding previously assumed rates of subclinical vasculitis.[^c15] Cardiac MRI detects subclinical myocardial fibrosis (elevated native T1) and impaired longitudinal strain in RA patients independent of traditional cardiovascular risk factors.[^c16] The 2025 ERS/EULAR guidelines recommend targeted HRCT screening for RA patients at risk of interstitial lung disease, and AI-based CT quantification has identified small airway loss as a potential imaging biomarker with AUC 0.801.
The EULAR-OMERACT combined ultrasound scoring system and the RAMRIS provide standardized, validated measures of inflammation and damage that are widely used as endpoints in clinical trials. Imaging-based treatment targets remain an area of active investigation: a randomized controlled trial comparing HandScan-guided treat-to-target with ACR/EULAR Boolean-remission-guided treatment in early RA found the HandScan target inferior, with HAQ 0.21 units worse at 18 months (95% CI 0.01–0.40) and high rates of perceived overtreatment.[^c19]