Related Experiment Video
Updated: Jun 13, 2026
Skin Biopsy for Diagnosing Discoid Lupus Erythematosus
Published on: June 10, 2025
Artificial intelligence for biopsies and imaging modalities in systemic autoimmune rheumatic diseases: An instructive
Konstantinos N Panagiotopoulos1, Nikos Tsiknakis2, Dimitrios I Zaridis3
1Department of Pathophysiology, Medical School, National and Kapodistrian University of Athens, 75 Mikras Asias Street, Goudi, 11527 Athens, Greece; Research Institute for Systemic Autoimmune Diseases, Athens, Greece.
Purpose:
To organize the existing literature regarding applications of artificial intelligence (AI) in biopsies and imaging modalities of patients with systemic autoimmune rheumatic diseases (SARDs) and to familiarize readers with the most commonly occurring concepts.
Results:
Firstly, we present a workflow that summarizes techniques implemented in AI for biopsies and imaging modalities in SARDs. Next, we describe challenges specific to image analysis for medicine. Subsequently, we describe the goals for an AI study in this field, and the prerequisites to meet them in SARDs. Finally, after reviewing the existing literature, we present the applications of AI for image analysis in each SARD. Accordingly, we analyze 1-2 studies from each SARD and mention key messages and lessons derived from them. Lastly, we create a recommendation landscape identifying unmet needs for AI applications in each SARD. The vast majority of studies employ supervised learning for image classification or segmentation, and rarely for regression. The median dataset size was 116 patients for imaging studies and 271 patients for biopsies studies, while the number of images per study varied greatly. Reporting of multiple performance metrics was frequently neglected.
Conclusions:
Employing AI for SARD image analysis ultimately demands large datasets with multimodal and adequately diverse data to effectively capture the heterogeneity of SARDs. In the field of rheumatology, plagued by subjectivity and interobserver variability, issues regarding data quality, regulatory authorities and the specificity and clinical impact of questions posed will define the time needed for clinical adoption of AI-assisted medical care.
