Artificial intelligence for acute appendicitis diagnosis: A systematic review of current evidence, challenges, and...

Acute appendicitis remains one of the most common surgical emergencies, yet its diagnosis continues to pose challenges due to overlapping clinical presentations and variable imaging findings. Artificial intelligence (AI) has recently emerged as a promising tool to enhance diagnostic accuracy, assist clinical decision-making, and potentially reduce negative appendectomy rates.

A systematic literature search was conducted in PubMed, Scopus, and Web of Science databases for English-language studies published between January 2015 and September 2025. Search terms included "artificial intelligence," "machine learning," "deep learning," and "acute appendicitis." Eligible studies focused on diagnostic model development or validation using clinical, laboratory, or imaging data. Extracted parameters included study design, AI algorithm, input modality, and diagnostic performance metrics (area under the curve [AUC], accuracy, sensitivity, specificity, and predictive values). Because of study heterogeneity, a qualitative synthesis was performed instead of a meta-analysis.

Clinical- and laboratory-based AI models frequently outperformed traditional scoring systems, demonstrating high negative predictive value and AUCs ranging from 0.85 to 0.93. Deep learning models applied to computed tomography imaging achieved the highest diagnostic accuracy, often exceeding junior radiologists, with reported AUCs up to 0.96. Ultrasound-based AI frameworks improved operator-independent detection, while magnetic resonance imaging-AI studies remain limited but promising. Hybrid multimodal models integrating clinical, laboratory, and imaging data achieved balanced performance (AUC ≈ 0.90-0.93) and hold the greatest translational potential. However, most studies were retrospective, single-center, with insufficient external validation and inconsistent methodological reporting.

AI demonstrates substantial potential to enhance diagnostic workflows for suspected appendicitis, particularly for identifying complicated cases and minimizing unnecessary imaging or surgery. Future research should emphasize multicenter prospective validation, standardized data annotation, explainable AI methods, and seamless integration into electronic health records and picture archiving systems.