Introduction: Liver surgery requires identification of tumor(s) in relation to key vessels to preserve healthy tissue, while obtaining negative margins. Current planning is mentally strenuous for residents as it relies on mental 3D reconstruction, anatomical knowledge, spatial sense, and experience. The purpose of this study is to determine if 3D segmentation improves resident ability to devise appropriate liver resection plans.

Methods: Senior general surgery residents were recruited. Pre-operative CT/MR images were selected if they reflected actual surgeries performed. Images were segmented to create interactive 3D models. Residents were asked to devise surgical plans for case-matched 2D and 3D models in an alternating, randomly generated order. Primary outcome was correct preoperative plan based on actual surgery performed. Secondary outcome was time (seconds(s)) to devise plan. Planning data was analyzed using Wilcoxon test, time was analyzed using paired t-test.

Results: All 14 senior residents from our institution participated. The average correct response was 1.7 of 5 (34%; range 1 to 4) for the 2D group, and 3.1 of 5 (62%; range 0 to 4) for the 3D group (p<0.01). The average time to complete each plan was 156±107s for the 2D group, and 84±73s for the 3D group (p<0.01).

Conclusions: The results show that 3D segmentation increases accuracy of surgical planning and decreases amount of time required. 3D segmentation is useful as a teaching tool as it reduces cognitive load required to mentally reconstruct 2D images, allowing the resident to focus on surgical planning. It improves understanding of spatial liver anatomy and serves as an adjunct to current 2D planning methods. This has the potential to be developed into a module for teaching liver surgery relevant to a competency-based curriculum.