Artificial intelligence and intra-abdominal infections: A systematic review of diagnostic

Abstract

Intra-abdominal infections (IAIs) are common and life-threatening complications of abdominal surgery, trauma, and acute abdomen, encompassing diverse etiologies such as peritonitis, appendicitis, cholangitis, diverticulitis, pancreatitis, and intra-abdominal abscesses. Timely and accurate diagnosis is pivotal for optimizing treatment strategies and improving patient prognosis, yet remains clinically challenging. With the rapid advancement of artificial intelligence (AI), innovative approaches have emerged to enhance early detection, etiological identification, and disease monitoring of IAI. This systematic review synthesizes evidence from 103 eligible studies to comprehensively evaluate the current state and advancements of AI applications in IAI management. Two core trends were identified: (1) development of AI models as alternatives to traditional scoring systems for IAI severity prediction; and (2) construction of etiology-focused predictive models for infection source identification. Key limitations of existing research include inadequate data quality (limited sample size, inconsistent collection standards, insufficient data diversity) and poor model generalization, which hinder clinical translation. This review further discusses the challenges and actionable directions for integrating AI into clinical practice, emphasizing the need for multicenter data sharing, standardized frameworks, and enhanced model interpretability. Collectively, AI holds substantial potential to revolutionize IAI diagnosis and management, particularly in multimodal data integration and real-time monitoring, but requires addressing infrastructural and regulatory barriers to achieve meaningful clinical adoption.