A Multi-Center Benchmark for Abdominal Disease Diagnosis and Report Generation from Non-Contrast CT

2026-06-15 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Clinical Care & Medical Practice, Health & Medical Research · Depth: Expert, quick

Summary

A novel multi-center benchmark has been introduced for multi-organ abdominal disease diagnosis and automated radiology report generation, aiming to synthesize contrast-enhanced findings from single-phase non-contrast CT (NCCT). This initiative addresses challenges associated with multiphasic contrast-enhanced CT (CECT), including contrast-induced nephropathy risks, increased acquisition burden, and radiologist workload. Researchers curated a large-scale dataset of paired NCCT-CECT studies and their corresponding contrast-enhanced radiology reports from two centers, divided into internal and external validation cohorts. Five contemporary deep learning architectures, spanning chest-specific, abdomen-specific, and general-purpose multimodal domains, were benchmarked under a unified evaluation protocol. Experiments demonstrated that NCCT retains diagnostic signals, achieving an average multi-organ AUC of 69.1% on the internal cohort and 63.1% on the external cohort. The public release of this dataset and standardized benchmark seeks to accelerate research into safer, resource-efficient, and globally accessible contrast-free abdominal imaging workflows.

Key takeaway

For AI Scientists and Research Scientists developing medical imaging solutions, this benchmark demonstrates that non-contrast CT can achieve significant diagnostic accuracy (AUC 69.1% internal, 63.1% external) for abdominal diseases. You should consider leveraging the publicly released dataset and benchmark to develop and validate contrast-free imaging models, potentially reducing patient risks and radiologist workload. This shifts focus towards synthesizing CECT findings from NCCT.

Key insights

Non-contrast CT retains diagnostic signals for abdominal disease, enabling safer, resource-efficient imaging workflows.

Principles

• Non-contrast CT can provide sufficient diagnostic signals for multi-organ abdominal disease.
• Synthesizing contrast-enhanced findings from NCCT reduces CECT risks and workload.

Method

A multi-center benchmark was established by curating paired NCCT-CECT studies and reports, then evaluating five deep learning architectures for multi-organ diagnosis and report generation.

In practice

• Access the publicly released dataset for developing contrast-free abdominal imaging models.
• Benchmark novel deep learning architectures against the established protocol.

Topics

• Abdominal Disease Diagnosis
• Non-Contrast CT
• Radiology Report Generation
• Deep Learning Benchmarks
• Medical Imaging Datasets
• Contrast-Free Imaging

Code references

• xmed-lab/TriALS-Report

Best for: AI Scientist, Research Scientist, Computer Vision Engineer

Related on AIssential

• Benchmarks and methods for 3D medical image retrieval — The 3D-MIR benchmark and methods advance medical image retrieval by providing standardized evaluation and multi-modal search strategies.
• Quantitative coronary calcification analysis for prediction of myocardial ischemia using non-contrast CT calcium scoring — Quantitative calcium-omics features from CTCS scans enhance myocardial ischemia prediction beyond traditional risk factors.
• Impact of Synthetic Lesional MR Images in Automated Focal Cortical Dysplasia Detection in Low-Data Scenarios — Synthetic FCD-MRIs generated by conditional networks can augment real data, reducing labeled data needs for automated detection.
• Visual Search Patterns in 3D Pancreatic Imaging: An Eye Tracking Study — This study investigates visual search patterns and strategies employed by radiologists when assessing the pancreas in 3D computed tomography (CT) images, addressing a gap in eye tracking research for…
• CA-GCL: Cross-Anatomy Global-Local Contrastive Learning for Robust 3D Medical Image Understanding — CA-GCL uses global-local contrastive learning and clinical-aware text augmentation to prevent representation collapse in 3D medical FVLP.
• Label-free pathological subtyping of non-small cell lung cancer using deep classification and virtual immunohistochemical staining — Label-free autofluorescence imaging combined with deep learning accurately subtypes NSCLC and generates virtual IHC stains.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.