Artificial Intelligence in Gallbladder Carcinoma: A Review of Research Studies and Clinical Evidence.
Authors
Affiliations (11)
Affiliations (11)
- Department of Colorectal Surgery, Imam Reza Hospital, AJA University of Medical Sciences, Tehran, Iran.
- Department of Internal Medicine, Yazd Branch, Islamic Azad University, Yazd, Iran. [email protected].
- Department of Internal Medicine, School of Medicine, Zanjan University of Medical Sciences, Zanjan, Iran.
- Department of General Surgery, School of Medicine, Babol University of Medical Sciences, Babol, Iran.
- Clinical Research Development Unit, Shahid Beheshti Hospital, Babol University of Medical Sciences, Babol, Iran.
- Breast Health & Cancer Research Center, Iran University of Medical Sciences, Tehran, Iran.
- Department of General Surgery, School of Medicine, Iran University of Medical Sciences, Tehran, Iran.
- Department of Surgery, School of Medicine, Iranshahr University of Medical Sciences, Iranshahr, Iran.
- Department of Medical Genetics, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran.
- Nanomedicine and Nanobiology Research Center, Shiraz University of Medical Sciences, Shiraz, Iran.
- Hematology and Oncology Research Center, Non-Communicable Diseases Research Institute, Shahid Sadoughi University of Medical Sciences, Yazd, Iran.
Abstract
Gallbladder carcinoma (GBC) is the most common biliary tract malignancy and is usually diagnosed at an advanced stage. Artificial intelligence (AI), including radiomics, machine learning (ML) and deep learning (DL), is increasingly applied to its diagnosis, surgical management and prognostication, but how far the reported performance can be trusted is unclear. PubMed/MEDLINE, Embase, Web of Science, the Cochrane Library and Google Scholar were searched to 5 May 2026 using database-specific search strategies. Each primary prediction-model study was characterized separately by design, number of centers and validation type (internal, temporal or external). Of 31 primary AI studies in GBC, 29 were retrospective, 22 were single-center and 24 reported internal validation only. In independent test or external cohorts, DL and ML models discriminated benign from malignant gallbladder lesions with areas under the curve (AUCs) of approximately 0.80-0.89; a multicenter model distinguishing xanthogranulomatous cholecystitis from GBC achieved an external AUC of 0.88, and automated computed tomography (CT) segmentation an external Dice score of 0.76. The best recurrence model achieved an AUC of 0.76 on internal (random-split) validation; none of the reviewed prognostic models reported a quantified external-validation estimate. Intraoperative AI is technically feasible but has no outcome data, and federated learning has been tested only in simulation. AI could plausibly assist several steps of GBC care, but the evidence is dominated by retrospective, single-center, internally validated studies from Asia, and metrics from different studies are not directly comparable. Prospective multicenter external validation, shared annotated datasets and standardized reporting are required before clinical adoption.