Methods and Reproducibility of Liver Fat Measurement Using 3-Dimensional Liver Segmentation From Noncontrast Computed...

Area of Science:

• Radiology
• Cardiology
• Hepatology

Background:

• Nonalcoholic fatty liver disease (NAFLD) shares risk factors with cardiovascular disease (CVD).
• NAFLD independently predicts increased CVD risk and adverse outcomes.
• Accurate and reproducible measurement of liver fat is crucial for understanding this relationship.

Purpose of the Study:

• To evaluate the reproducibility of a 3D liver volume segmentation method for diagnosing fatty liver on noncontrast cardiac CT.
• To compare this 3D method with established 2D segmentation criteria for liver fat measurement.

Main Methods:

• The study included 68 participants from the EVAPORATE trial undergoing serial noncontrast cardiac CT.
• Fatty liver was diagnosed using liver attenuation < 40 Hounsfield units (HU).
• Both 2D and 3D liver segmentation were performed using Philips software, with Bland-Altman analysis for reproducibility.

Main Results:

• Interreader reproducibility for 3D liver mean HU measurements was 96%.
• Reproducibility for 2D and 3D liver mean HU measurements was 93%.
• Agreement (Kappa) between 2D and 3D methods in identifying fatty liver was excellent at 96.4%.

Conclusions:

• 3D liver segmentation on noncontrast cardiac CT offers a reliable and reproducible method for diagnosing and serially measuring fatty liver.
• This 3D approach can be a valuable imaging biomarker for exploring links between atherosclerosis, fatty liver, and CVD risk.
• Further research is needed to assess the sensitivity of 3D segmentation for low liver fat content compared to 2D methods.