ARRS 2022 Abstracts

RETURN TO ABSTRACT LISTING


2046. Comparing Thin and Thick Section Abdominal CT Image Data for Fully Automated AI-Based Body Composition Algorithms
Authors * Denotes Presenting Author
  1. Matthew Lee *; University of Wisconsin
  2. John Garrett; University of Wisconsin
  3. Alberto Perez; University of Wisconsin
  4. Ronald Summers; National Institutes of Health Clinical Center
  5. Perry Pickhardt; University of Wisconsin
Objective:
The purpose of this study is to compare biometric measures derived from thin (1.25-mm) section abdominal CT source image data with thick (5-mm) sections for generating body composition data using fully automated algorithms.

Materials and Methods:
In this retrospective study, fully automated CT-based body composition algorithms for quantifying visceral-to-subcutaneous (V/S) fat ratio, muscle attenuation, muscle area, aortic calcification, liver attenuation, liver volume, and spleen volume were applied to both thin (1.25-mm) and thick (5-mm) abdominal CT series from two patient cohorts: unenhanced scans of asymptomatic adults undergoing colorectal cancer screening, and post-contrast scans of patients with colorectal cancer. Biometric measures derived from the thin and thick section data were compared, including correlation coefficients and Bland-Altman analysis.

Results:
A total of 9882 CT scans (mean age, 57.0 years; 4527 women, 5355 men) were evaluated, including 8947 non-contrast and 935 contrast-enhanced CT examinations. Strong correlation was observed for all biometric measures: V/S fat ratio (r=0.99, r2=0.97), muscle attenuation (r=0.99, r2=0.97), muscle area (r=0.99, r2=0.98), aortic calcium (r=0.96, r2=0.93), liver attenuation (r=0.99, r2=0.98), liver volume (r=0.99, r2=0.98), and spleen volume (r=1.0, r2=1.0) (p<0.001 for all). Bland-Altman analysis showed strong agreement for muscle attenuation, muscle area, liver attenuation, liver volume, and spleen volume. Mean percentage differences amongst body composition metrics were less than 5% for V/S ratio (4.6%), muscle area (-0.5%), liver attenuation (0.4%), and liver volume (2.7%), and less than 10% for muscle density (-5.5%) and spleen volume (5.1%). For aortic calcium, Agatston scores agreed for mild (0-100) and severe (>400) burden in 91.3% and 92.3% of cases, respectively. No significant agreement differences were observed between the non-contrast and post-contrast cohorts.

Conclusion:
Automated biometric measures derived from thin and thick abdominal CT data are strongly correlated and show agreement, particularly for soft tissue applications, making it feasible to use either series for these CT-based body composition algorithms for the purposes of opportunistic screening.