2407. ACR Accreditation Results Using CT Phantoms: Implications for Opportunistic CT Screening of Sarcopenia, Adiposity, and Emphysema
Authors * Denotes Presenting Author
  1. Robert Boutin *; Stanford University School of Medicine
  2. Andrew Hernandez; UC Davis School of Medicine
  3. Leon Lenchik; Wake Forest School of Medicine
  4. J. Anthony Seibert; UC Davis School of Medicine
  5. Dustin Gress; American College of Radiology
To investigate if there is a systematic bias in CT number (Hounsfield unit, HU) measurements relevant to quantitative opportunistic CT evaluation of muscle (for myosteatosis in sarcopenia), adipose tissue (for degree of adiposity), and lung (for pulmonary bullae in emphysema) by comparing scanners made by four major CT manufacturers.

Materials and Methods:
CT number accuracy data acquired using the American College of Radiology (ACR) accreditation phantom were evaluated in a blinded fashion for four CT manufacturers (A: n=8,500; B: n=18,575; C: n=8,278, D: n=32,039) using an adult abdomen CT technique (120 kV, 240 mA, 50-cm FOV, standard reconstruction algorithm). Analysis of the CT number accuracy module of the ACR accreditation phantom evaluated the following reference standard materials: acrylic (120 HU), water (0 HU), polyethylene (-95 HU, surrogate for adipose tissue), and air (-1000 HU). A 1.5 interquartile range outlier removal method was used as a robust method for extreme outlier removal with a nonnormal distribution, resulting in removal of 2.8% of examinations averaged across all manufacturer/material combinations. Differences in HU values across all manufacturers were assessed using the Kruskal-Wallis test followed by a post-hoc test for pairwise comparisons.

For skeletal muscle (with a cutpoint for diagnosing myosteatosis typically ~30-40 HU), the difference in mean CT numbers among manufacturers was between acrylic (up to 6.1 HU, between manufacturers A and C) and water (up to 3.0 HU between manufacturers B and D). For these materials, the differences in HU values between the four CT manufacturers were highly significant (p < 0.001). For the adipose tissue surrogate (polyethylene), differences in CT numbers across all manufacturers were also highly significant (p < 0.001), with mean values [standard deviation, SD] of -94.5 [3.6], -89.4 [3.6], -87.4 [3.4], and -89.6 [3.5] for manufacturers A, B, C, and D, respectively. The mean CT numbers of air across all manufacturers were also highly significant (p < 0.001), with mean values [SD] of -991.7 [10.1], -992.5 [4.4], -987.7 [5.3], and -979.0 [8.5] for manufacturers A, B, C, and D, respectively.

CT number measurements between manufacturers have a systematic offset when compared to each other, and these offsets change with the tissue measured. Knowledge of these offsets may be useful in understanding the limitations and potential for opportunistic quantitative CT diagnosis of sarcopenia, adiposity, and emphysema.