Fig 1.
Flow chart describing pooled cohorts and scan cohort scan phase heterogeneity.
Table 1.
Patient demographics, scan acquisitions, and vertebral levels within scan limits.
Fig 2.
Aorta attenuation (AoHU) is characterized by the mean and standard deviation of pixels inside a central aorta region.
Fig 3.
Example pixel misclassification types when detecting calcified plaque via HU thresholds.
Contrast-enhanced scans have higher baseline HU within the aorta resulting in false positive errors at low thresholds. Higher HU thresholds reduce false positive errors but can produce false negative errors for plaque pixels of moderate density.
Fig 4.
Distribution of mean AoHU at L1 for males and females within each contrast enhancement group.
Table 2.
Simple linear regression coefficient (Est.) and fit statistic (R2) for mean AoHU versus demographic and scan factors.
Fig 5.
Distribution of standard deviation in AoHU within the center of the aorta at L1 for males and females within each contrast enhancement group.
Table 3.
Changes in standard deviation AoHU by demographic factors.
Fig 6.
Distribution (bars) and cumulative density function (line) of calcified plaque pixel attenuation in 50 non-contrast CT scans.
The CDF line reflects the false-negative detection rate as true plaque pixels are excluded at increased CT thresholds.
Fig 7.
Estimated rates of false-positive (aorta contents incorrectly identified as plaque) and false-negative (plaque content incorrectly identified as aorta) errors when thresholding pixels within the aorta to detect calcified plaque regions.
A uniform threshold scheme shares the same single threshold across all scans and slices. A dynamic scheme calculates separate thresholds based on pixels observed within the central aorta.