Technique automates CTA cerebral artery segmentation
CT angiography (CTA) is rapidly becoming the preferred diagnostic imaging exam of choice for vascular imaging in many applications. Although digital subtraction angiography (DSA) has long been the reference standard for vascular imaging, it is an invasive and costly procedure requiring high contrast doses, which carries with it a risk for complications.

However, the utilization of CTA poses challenges for diagnostic imaging; primarily the interpretation of the large data sets acquired during these procedures. As such, the utilization of automated, advanced visualization technology for the detection and quantification of vascular disease is of great interest by interpreting clinicians.

A research team from the departments of medical informatics and radiology at Erasmus Medical Center in Rotterdam, the Netherlands, and the University Medical Center Utrecht in Utrecht, the Netherlands, has developed technology that allows for the fully automated segmentation of cerebral arteries in CTA without an additional CT scan for bone suppression.

The results of their application of this technique were published in this month’s Radiology.

“The purpose of our study was to retrospectively assess the feasibility of a fully automated image post-processing tool for the segmentation of arterial cerebrovasculature from CTA,” they wrote.

CTA data were acquired from 27 patients who had exams conducted with a Philips Healthcare 16-slice Mx8000 LDT CT system from February 2003 to March 2004. All patients presented with acute subarachnoid hemorrhage and underwent a CTA as part of their diagnostic workup. Nonionic contrast, Ultravist by Schering, was employed during the exams, which took approximately 2 minutes.

Sagittal view of a volume-rendered cerebral CT angiographic data set image. Image and caption courtesy of the Radiological Society of North America.  
The segmentation method utilized by the researchers was conducted in five steps on images from 18 patients in the original data set.

First, regions of interest are automatically determined on the basis of anatomic information contained in the patient data set. Second, non-arterial structures are suppressed as much as possible without affecting the arterial vasculature.

Next, seed points were places in the feeding arteries of the circle of Willis by searching for circular structures in the CT cross-sectional images. The fourth step was a rough segmentation of the arterial vasculature achieved by combining segmentations of internal carotid arteries and the intracranial vasculature. The final step used the rough segmentation as an initialization of another computer algorithm to achieve accurate vessel boundary localization.

A statistical analysis conducted on the method found a 95 percent confidence interval of the success rate. Overall, they reported the vessel segmentation method had a success rate of 83 percent (15 of 18 patients).

“Failure was mainly caused by the absence of a circular shape of the vessel cross sections,” they wrote. “The failures were manually corrected in the axial sections.”

One of the most important attributes of the technique developed by the research team is that it does not require an additional scan for bone masking and thus avoids additional radiation exposure to the patient.

“We believe that the proposed segmentation method may provide a basis for the development of advanced clinical applications, which includes detection, characterization, and quantification of plaques, calcifications, stenoses, aneurysms, and vasospasm,” the authors wrote.