Local, semi-automatic, three-dimensional liver reconstruction or external provider? : An analysis of performance and time expense

Abstract

PURPOSE In hepatobiliary surgery, preoperative three-dimensional reconstruction based on CT or MRI can be provided externally or by local, semi-automatic software. We analyzed the time expense and quality of external versus local three-dimensional reconstructions.

METHODS Three first-year residents reconstructed data from 20 patients with liver pathologies using a local, semi-automatic, server-based program. Initially, five randomly selected patient datasets were segmented, with the visualization of an established external company available for comparison at all times (learning phase). The other fifteen cases were compared with the external datasets after completing local reconstruction (control phase). Total time expense/case and for specific manual and semi-automated reconstruction steps were recorded. Segmentation quality was analyzed by testing the equivalence for liver and tumor volumes, portal vein sectors, and hepatic vein territories.

RESULTS The median total reconstruction time was reduced from 2.5 h (learning phase) to 1.5 h (control phase) (− 42%; p < 0.001). Comparing the total and detailed liver volumes (sectors and territories) as well as the tumor volumes in the control phase equivalence was proven. In addition, a highly significant correlation between the external and local analysis was obtained over all analyzed segments with a very high ICC (median [IQR]: 0.98 [0.97; 0.99]; p < 0.01).

CONCLUSION Local, semi-automatic reconstruction performed by inexperienced residents was feasible with an expert level time expense and the quality of the three-dimensional images was comparable with those from an external provider.