Development and Validation of a Fast Volumetric Determination Method of Visceral Adipose Using CT Images
Z Yu1*, A Parikh2, A Klopp1, J Cheung1, L Court1, (1) MD Anderson Cancer Center, Houston, TX, (2) Rice University, Houston, TXSU-C-134-7 Sunday 1:00PM - 1:55PM Room: 134
Purpose: To develop and validate a method for quick and accurate calculation of visceral adipose volume with minimal user interaction. The goal is to provide a tool that can be reliably used to analyze sizable patient databases.
Methods: CT images are imported into in-house software. Superior and inferior boundaries of the visceral region are defined by the user. Five ellipses, spaced equal distance apart between these boundaries, are then manually drawn on axial slices inside the rectus abdominis and transverse abdominis muscles to define the border between visceral and subcutaneous adipose. Ellipses are automatically interpolated to fill the rest of the region. A histogram of Hounsfield Units of all voxels within the visceral region is obtained. The threshold for adipose was defined to be the peak adipose value ±1.25 FWHM. The accuracy of the ellipse estimation method was validated by taking ten patient CT images, 5 obese and 5 lean, and comparing them with manually delineated rectus abdominis and transverse abdominis muscles. The accuracy of visceral fat measurement was validated using a CT scan of a physical phantom which was made using animal fat (to simulate human adipose) and other animal tissues.
Results: The mean difference between manually delineated and ellipse-based visceral adipose volumes for all ten patients was -1.3%. For nine out of ten patients, this difference was less than 5%. For the phantom measurements, our calculations using CT images differed from absolute fat volume by less than 10% for all scanning parameters. Accuracy of volume determination is inversely proportional to slice thickness and field-of-view, and not affected by kVp or mAs.
Conclusion: We developed and validated a method to determine the total volume of visceral adipose based on CT imaging. The entire process takes on average 2 minutes per patient and will be useful for large population studies.