Radiology: Promising processing tool automates stenosis measurements
A virtual elastic sphere processing tool offers a quantifiable, reproducible mechanism to measure vessel stenosis on CT angiography and MR angiography datasets, according to a study published online July 25 in Radiology.
Although carotid endarterectomy can reduce risk of subsequent stroke for patients with transient ischemic attacks and carotid artery stenosis, the benefit of the procedure hinges on the degree of stenosis. However, current methods for measuring stenosis are manual, reader-dependent and minimally reproducible.
Given the situation, researchers from Helmholtz Institute of Biomedical Engineering, University of Aachen in Aachen, Germany sought to develop and evaluate a tool for efficient, accurate and reproducible quantification of blood vessel stenosis in CT angiographic images, using a dual-source configuration, and MR angiographic images.
To use the tool, a physician determines start and end (seed) points in a 3D dataset. Then a virtual elastic sphere is fitted through the segmented vessel lumen and the diameter is recorded, explained Felix Gremse, Dipl-Inf, University of Aachen, and colleagues. A 3D distance map traces the path of the sphere to produce the measurement.
Gremse and colleagues evaluated the virtual elastic sphere method using a three-tiered method. First, they measured stenosis with a virtual caliper and then using the seed points, they conducted CT exams of stenosis phantoms and imaged mice via micro-CT to establish ground truth. Finally, two radiologists blinded to each other’s measurements analyzed CT and MR patient datasets of 16 clinically symptomatic patients with carotid artery stenosis between May 2007 and April 2008.
The researchers reported minimal lumen diameters from the phantoms correlated strongly with micro-CT datasets used as ground truth values. Progressing to patient data, the interreader correlation was significantly stronger for automated segmentation versus manual segmentation.
Furthermore, CT, dual-energy CT and MR offered high sensitivity (87 percent, 93 percent and 83 percent, respectively) and specificity (100 percent, 100 percent and 90 percent, respectively).
The virtual sphere method also improved efficiency, dropping the mean reader time from 104 +/- 27 seconds per carotid artery with manual measurement to 35 +/- 8 seconds per carotid artery with the automated system.
The system improves on earlier attempts to automate stenosis measurements, as these mechanisms did not improve or evaluate reproducibility, explained the authors. Gremse and colleagues acknowledged that the lack of catheter-based angiography in the patient group was a limitation as such data could have determined ground truth stenosis scores.
Nevertheless, the researchers claimed the proof of concept a success. “The virtual elastic sphere tool is easy to use, time-efficient, and improves reproducibility over that achieved with manual stenosis measurements,” Gremse et al concluded. In addition, they noted that the tool may be applied to patients with other conditions such as stenosis of the aorta or iliac arteries, or to determine the optimal size of stents.
Although carotid endarterectomy can reduce risk of subsequent stroke for patients with transient ischemic attacks and carotid artery stenosis, the benefit of the procedure hinges on the degree of stenosis. However, current methods for measuring stenosis are manual, reader-dependent and minimally reproducible.
Given the situation, researchers from Helmholtz Institute of Biomedical Engineering, University of Aachen in Aachen, Germany sought to develop and evaluate a tool for efficient, accurate and reproducible quantification of blood vessel stenosis in CT angiographic images, using a dual-source configuration, and MR angiographic images.
To use the tool, a physician determines start and end (seed) points in a 3D dataset. Then a virtual elastic sphere is fitted through the segmented vessel lumen and the diameter is recorded, explained Felix Gremse, Dipl-Inf, University of Aachen, and colleagues. A 3D distance map traces the path of the sphere to produce the measurement.
Gremse and colleagues evaluated the virtual elastic sphere method using a three-tiered method. First, they measured stenosis with a virtual caliper and then using the seed points, they conducted CT exams of stenosis phantoms and imaged mice via micro-CT to establish ground truth. Finally, two radiologists blinded to each other’s measurements analyzed CT and MR patient datasets of 16 clinically symptomatic patients with carotid artery stenosis between May 2007 and April 2008.
The researchers reported minimal lumen diameters from the phantoms correlated strongly with micro-CT datasets used as ground truth values. Progressing to patient data, the interreader correlation was significantly stronger for automated segmentation versus manual segmentation.
Furthermore, CT, dual-energy CT and MR offered high sensitivity (87 percent, 93 percent and 83 percent, respectively) and specificity (100 percent, 100 percent and 90 percent, respectively).
The virtual sphere method also improved efficiency, dropping the mean reader time from 104 +/- 27 seconds per carotid artery with manual measurement to 35 +/- 8 seconds per carotid artery with the automated system.
The system improves on earlier attempts to automate stenosis measurements, as these mechanisms did not improve or evaluate reproducibility, explained the authors. Gremse and colleagues acknowledged that the lack of catheter-based angiography in the patient group was a limitation as such data could have determined ground truth stenosis scores.
Nevertheless, the researchers claimed the proof of concept a success. “The virtual elastic sphere tool is easy to use, time-efficient, and improves reproducibility over that achieved with manual stenosis measurements,” Gremse et al concluded. In addition, they noted that the tool may be applied to patients with other conditions such as stenosis of the aorta or iliac arteries, or to determine the optimal size of stents.