Threshold selection for CT angiography shaded surface display of the renal arteries

Abstract

The purpose of this study was to define objective criteria to calculate a tissue segmentation threshoid for shaded surface display {SSD) rendering of the renal arteries with computed tomography angiography. Contrast-enhanced spiral CT scans were obtained through the renal arteries of nine patients. Six sets of SSD images were rendered for each patient with Iower threshold values ranging from 80 to 130 Hounsfield units (HU) by increments of 10 HU. Visceral organ enhancement was measured in the aorta, liver, spleen, pancreas, and kidney. The segmentation threshold for each patient was determined by evaluation of the SSD images alone as well as by comparison with conventional arteriograms. The ideal threshold, as shown by comparison with conventional arteriography, was better correlated with a threshold value selected by qualitative evaluation of SSD images alone (rs = .42}, than with measured enhancement in visceral organs (rs = -.289 to .009). The degree of stenosis was overestimated in a single renal artery (1 of 18) because of an inappropriate threshold selected by evaluation of the SSD images alone. In comparison with a segmentation threshold calculated from measured enhancement of visceral organs, a segmentation threshold selected by qualitative evaluation of the resulting SSD images is more likely to approximate the ideal threshold. Given the subjective nature of such threshold selection, further evaluation is warranted to determine whether threshold selection may result in inaccurate grading of stenosis.

Volume Rendering versus Maximum Intensity Projection in CT Angiography: What Works Best, When, and Why

“…The threshold that would approximate the attenuation of bone in a healthy patient, for instance, exceeds the attenuation value for markedly osteopenic bone, a situation that may create gaps in the CT data and inaccuracies in the reconstructed image. Similarly, in CT angiography, incorrect threshold selection may result in inaccurate grading of vascular stenosis (7)(8)(9). The thresholding technique is also susceptible to noise introduced during scanning.…”

Segmentation of arteries in MPRAGE images of the ventral medial prefrontal cortex

“…Many vessel segmentation algorithms involve finding centerlines, or skeletons, of the vessels [23][24][25][26]. These algorithms are faster than nonskeleton algorithms, which work without finding a centerline [27][28][29]. None of these methods have been applied and rigorously tested for specific cortical regions, such as the VMPFC.…”