Related Experiment Video
Updated: Aug 6, 2026
Novel In Vivo Micro-Computed Tomography Imaging Techniques for Assessing the Progression of Non-Alcoholic Fatty Liver Disease
Published on: March 24, 2023
Automated Deep Learning Detection of Hepatic Steatosis on Noncontrast Computed Tomography Scans and Discrepancy With
David Yardeni1, Jianfei Liu2, Pritam Mukherjee2
1Liver and Energy Metabolism Section, Liver Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, Maryland; Department of Gastroenterology and Liver Diseases, Soroka University Medical Center, Beersheba, Israel.
Incidental hepatic steatosis on non-contrast CT scans is often unreported. A deep learning algorithm identified steatosis in 13.1% of scans, but radiologists only reported it in 32.7% of cases, highlighting a need for improved detection and reporting.
Area of Science:
- Radiology
- Hepatology
- Artificial Intelligence
Background:
- Metabolic dysfunction-associated steatotic liver disease (MASLD) is a growing global health concern.
- MASLD is often asymptomatic and undiagnosed, presenting a diagnostic challenge.
- Non-contrast CT (NCCT) can incidentally detect hepatic steatosis, but reporting rates are unknown.
Purpose of the Study:
- To determine the reporting rate of incidental hepatic steatosis found on NCCT scans.
- To investigate the utility of a deep learning algorithm for automated detection of hepatic steatosis.
- To identify factors associated with unreported hepatic steatosis on NCCT.
Main Methods:
- Retrospective analysis of 3,646 NCCT scans from adult patients (2012-20).
- Automated deep learning algorithm used for liver segmentation and attenuation assessment.
- Image-based steatosis defined as mean hepatic attenuation < 40 HU; findings compared to radiology reports.
Main Results:
- Hepatic steatosis was identified in 13.1% (480/3,646) of NCCT scans.
- Radiologists reported steatosis in only 32.7% (157/480) of identified cases.
- Unreported steatosis was associated with higher liver attenuation, varied fat distribution, and lower BMI.
Conclusions:
- Incidental hepatic steatosis on NCCT is frequently underreported.
- AI-based hepatic attenuation measurement could enhance steatosis detection and reporting rates.
- Improved reporting of incidental findings may aid in earlier MASLD diagnosis and management.
