Automated Vessel Segmentation in Dual Energy Computed Tomography Data of the Pelvis and Lower Extremities.

AIM: To evaluate the clinical feasibility of a newly developed, fully automatic vessel segmentation software with automatic structured bone elimination (ASBE) using graph-matching and subvoxel analysis.

MATERIALS AND METHODS: Dual energy computed tomography angiography (DECTA) data of 108 vessel segments were evaluated using the ASBE software and a commercial software against the digital subtraction angiography (DSA) standard of reference.

RESULTS: Using the ASBE software, sensitivity increased from 87.1% to 96.8% and data concordance with DSA increased from 64.5% to 88.6%, whereas specificity slightly decreased (79.2% vs. 87%) compared to the commercial software. Data concordance between ASBE software and DSA was especially high in severely stenosed (grade of stenosis >75%) blood vessels.

CONCLUSION: ASBE showed good concordance with the DSA standard of reference and non-inferiority compared to the commercial segmentation software. The main advantage of the ASBE software lies in its full automation and, thus, lower susceptibility to user prone errors.