Related Experiment Video
Updated: Jan 15, 2026
08:58
Artificial Intelligence Approaches to Assessing Primary Cilia
Published on: May 1, 2021
4.1K
Pilot Study of AI-Assisted ANA Immunofluorescence Reading-Comparison with Classical Visual Interpretation
Sarah Mayr1, Margit Dollinger1, Boris Ehrenstein1
1Department of Rheumatology and Clinical Immunology, Asklepios Medical Center Bad Abbach, 93077 Bad Abbach, Germany.
Journal of Clinical Medicine
|October 16, 2025
Summary
Artificial intelligence (AI) shows good agreement in detecting antinuclear antibodies (ANAs) for diagnosing rheumatic diseases. Further AI refinement is needed for reliable clinical screening of ANA patterns and titers.
Area of Science:
- Immunology
- Rheumatology
- Medical Diagnostics
Background:
- Antinuclear antibodies (ANAs) are key biomarkers for systemic autoimmune rheumatic diseases like lupus erythematosus.
- Indirect immunofluorescence testing (IIFT) with HEp-2 cells is the gold standard for ANA detection.
- Visual interpretation (VI) of IIFT is laborious; AI-based automated systems offer potential efficiency gains.
Purpose of the Study:
- To evaluate the diagnostic performance of an AI-based interpretation system (akiron® NEO) compared to visual interpretation (VI) for ANA detection.
- To assess the agreement in ANA positive/negative discrimination, titer levels, and pattern recognition between AI and VI methods.
- To determine the potential of AI-assisted IIFT as a reliable screening tool in clinical rheumatology.
Main Methods:
- 143 consecutive serum samples were analyzed using IIFT with both visual interpretation and the akiron® NEO AI system.
- ANA detection, titer levels, and patterns were compared between the two methods.
- Statistical analysis included Cohen's kappa (κ) for agreement and McNemar test for significant differences.
Main Results:
- Good agreement (κ=0.69) for positive/negative ANA discrimination at a cut-off of 80, improving to κ=0.76 at ≥1/80 with non-significant differences.
- Moderate agreement (κ=0.54) for ANA pattern recognition, improving to very good (κ=0.80) when specific patterns were grouped.
- AI-aided interpretation frequently assigned higher titer levels than visual interpretation.
Conclusions:
- AI-based interpretation demonstrates good agreement for ANA positive/negative discrimination and moderate to very good agreement for pattern recognition.
- The akiron® NEO system shows promise as a potential screening tool, but requires further algorithm refinement for pattern recognition and titer calibration.
- Continued research and development are essential for the successful clinical implementation of AI in ANA testing.
