Background: The aim of the present study was to evaluate the accuracy of two approaches using magnetic resonance imaging (MRI) or combined ultrasonography (US) and anti-cyclic citrullinated peptide antibody (ACPA) for diagnosis and classification of individuals with established rheumatoid arthritis (RA). Methods. In 53 individuals from a population-based, cross-sectional study, historic fulfilment of the American College of Rheumatology (ACR) 1987 criteria ("classification") or RA diagnosed by a rheumatologist ("diagnosis") were used as standard references. The sensitivity, specificity and Area under Curve for Receiver Operating Characteristics curves (ROC-area: (sensitivity + specificity)/2) were calculated for "current fulfilment of the ACR 1987 criteria" (list format), "adapted ACR 1987 criteria" (list format, substituting IgM rheumatoid factor with ACPA and clinical joint swelling and erosions on radiography with synovitis and erosions detected by US on a semi-quantitative scale), and RA MRI scoring System (RAMRIS) scores on low-field MRI in the unilateral hand. Results: For the ACR 1987 criteria the ROC-area was 75% (sensitivity/specificity = 50%/100%) (with "classification" as standard reference) and 69% (44%/94%) (with "diagnosis" as standard reference), while for the adapted ACR 1987 criteria it was 86% (75%/97%) (classification) and 82% (72%/91%) (diagnosis). For RAMRIS synovitis score in metacarpophalangeal (MCP) joints only (cut-off ≥5), the ROC-area (sensitivity/specificity) was 78% (62%/94%) (classification) and 85% (69%/100%) (diagnosis), while for the total synovitis score of MCP joints plus wrist (cut-off ≥10) it was 78% (62%/94%) (both classification and diagnosis). Conclusions: Compared with the ACR 1987 criteria, low-field MRI alone or adapted criteria incorporating US and ACPA increased the correct classification and diagnosis of RA.