State of Abdominal CT Datasets: A Critical Review of Bias, Clinical Relevance, and Real-world Applicability

2025-08-19

AI Summary

This study identifies critical limitations hindering the clinical deployment of abdominal CT-based AI models, including pervasive data redundancy (59.1% duplicate cases), strong geographic bias (75.3% from North America and Europe), domain shift (63%), and selection bias (57%)—all severely compromising model generalizability and fairness in resource-constrained settings. Method: We systematically evaluated 46 publicly available abdominal CT datasets using a novel, integrated assessment framework combining systematic review, a multidimensional bias evaluation schema, and quantitative diversity analysis. Contribution/Results: This work delivers the first large-scale, structured quality audit of the abdominal CT data ecosystem. It reveals systemic deficiencies undermining clinical robustness and geographic portability. Based on empirical findings, we propose a principled improvement paradigm centered on multi-center collaboration, standardized acquisition protocols, and inclusive coverage of diverse populations and imaging hardware. This provides both a methodological foundation and actionable pathways for developing clinically robust, geographically generalizable, and equitably deployable medical AI models.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This systematic review critically evaluates publicly available abdominal CT datasets and their suitability for artificial intelligence (AI) applications in clinical settings. We examined 46 publicly available abdominal CT datasets (50,256 studies). Across all 46 datasets, we found substantial redundancy (59.1% case reuse) and a Western/geographic skew (75.3% from North America and Europe). A bias assessment was performed on the 19 datasets with >=100 cases; within this subset, the most prevalent high-risk categories were domain shift (63%) and selection bias (57%), both of which may undermine model generalizability across diverse healthcare environments -- particularly in resource-limited settings. To address these challenges, we propose targeted strategies for dataset improvement, including multi-institutional collaboration, adoption of standardized protocols, and deliberate inclusion of diverse patient populations and imaging technologies. These efforts are crucial in supporting the development of more equitable and clinically robust AI models for abdominal imaging.

Problem

• Assessing bias and redundancy in abdominal CT datasets
• Evaluating geographic and clinical diversity limitations
• Addressing dataset issues to improve AI generalizability

Innovation

• Multi-institutional collaboration for dataset diversity
• Standardized protocols to reduce bias
• Inclusion of diverse patient populations