RSNA 2007 

Abstract Archives of the RSNA, 2007


SST08-04

Accuracy of Quantitative Vascular Measurements Using Automatic Vessel Tracking and Segmentation Software and Its Dependence on Dose and Filtration

Scientific Papers

Presented on November 30, 2007
Presented as part of SST08: Vascular/Interventional (Vascular Imaging—CTA)

Participants

Srinivas Aluri PhD, Abstract Co-Author: Nothing to Disclose
Uri Shreter, Presenter: Employee, GE Healthcare, Waukesha, WI
Tiffany Dugal, Abstract Co-Author: Nothing to Disclose
Stanley Fox, Abstract Co-Author: Nothing to Disclose

PURPOSE

We are evaluating a new software suite developed to automatically track vasculature. We use images from a vascular phantom scan to show that the software can track and measure vasculature 2 mm or larger in diameter.

METHOD AND MATERIALS

We scanned a vascular phantom with inserts simulating vasculature and stenosis repeatedly using a GE LightSpeed VCT scanner. Scanning was performed with standard and sharp kernels, using mA range from 10 to 335. The images were then read on an Advantage Windows workstation using VesselIQ Xpress software(GE). The software was used to track vessels between two points. The software automatically segments and calculates diameters and areas on the tracked vessels. Measurements were taken along the vessel at random intervals. The minimum diameter at the stenosis was measured.

RESULTS

Noise was measured in a flat field area of the phantom. The noise levels varied between 4.06(+/-0.27) and 42.78(+/-2.47) for standard kernel and 13.88(+/-1.27) and 149.9(+/-19.61) for sharp kernel. The software was able to reliably track the vessels 2mm and above across all noise levels and 1mm vessel at lower noise levels. The vessel diameter and stenosis measurements (in mm) show no significant difference with reduction of dose. The 2 mm vessel diameter at lowest noise levels were (1.99(+/-0.10) and 1.96(+/-0.11)) with standard and sharp kernels. At the noise level of about 40, the diameters were (2.13(+/-0.23) and 1.94(+/-0.18)) with standard and sharp kernels. Three different stenoses were measured (1 mm, 2 mm and 3 mm obstructions in a 4mm vessel) and the values were 2.9(+/-0.09), 1.99(+/-0.09) and 1.42(+/-0.17). The resolution of the scanner is the limiting factor in measuring the 1mm vessel and stenosis.

CONCLUSION

The VesselIQ Xpress software can accurately and reliably track, segment and measure vessels of diameters 2mm or higher, even with low dose and in the presence of image noise. Diameter and stenosis measurement accuracy better than 12% may be achieved at typical noise levels for these vessel sizes.

CLINICAL RELEVANCE/APPLICATION

The ability of VesselIQ Xpress software to perform vessel analysis in exams with lower dose can result in significant dose reduction in routine contrast enhanced cardiac CT examinations.

Cite This Abstract

Aluri, S, Shreter, U, Dugal, T, Fox, S, Accuracy of Quantitative Vascular Measurements Using Automatic Vessel Tracking and Segmentation Software and Its Dependence on Dose and Filtration.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5013496.html