Semiautomatic Cone-Beam Computed Tomography Virtual Hepatic Volumetry for Intra-Arterial Therapies

Purpose: To evaluate a software simulating the perfused liver volume from virtual selected embolization points on proximal enhanced cone-beam computed tomography (CT) liver angiography data set using selective cone-beam CT as a reference standard. Materials and Methods: Seventy-eight selective/proximal cone-beam CT couples in 46 patients referred for intra-arterial liver treatment at 2 recruiting centers were retrospectively included. A reference selective volume (RSV) was calculated from the selective cone-beam CT by manual segmentation and was used as a reference standard. The virtual perfusion volume (VPV) was then obtained using Liver ASSIST Virtual Parenchyma software on proximal cone-beam CT angiography using the same injection point as for selective cone-beam CT. RSV and VPV were then compared as absolute, relative, and signed volumetric errors (ABS_Err, RV_Err, and SV_Err, respectively), whereas their spatial correspondence was assessed using the Dice similarity coefficient. Results: The software was technically successful in automatically computing VPV in 74 of 78 (94.8%) cases. In the 74 analyzed couples, the median RSV was not significantly different from the median VPV (394 mL [196–640 mL] and 391 mL [192–620 mL], respectively; P = .435). The median ABS_Err, RV_Err, SV_Err, and Dice similarity coefficient were 40.9 mL (19.9–97.7 mL), 12.8% (5%–22%), 9.9 mL (−49.0 to 40.4 mL), and 80% (76%–84%), respectively. No significant ABS_Err, RV_Err, SV_Err, and Dice similarity coefficient differences were found between the 2 centers (P = .574, P = .612, P = .416, and P = .674, respectively). Conclusions: Perfusion hepatic volumes simulated on proximal enhanced cone-beam CT using the virtual parenchyma software are numerically and spatially similar to those manually obtained on selective cone-beam CT.