Evaluation of operator variability and technical accuracy of automatic image-based registration in liver fusion imaging.

Purpose: To evaluate the technical feasibility and performance of automatic image-based registration (IBR) for liver fusion imaging and to identify clinical and anatomical factors affecting registration success.

Materials and methods: This prospective study included 84 patients undergoing liver fusion imaging using an ultrasound system with IBR. Three operators with 5, 10, and 25 years of experience (junior, intermediate, and senior), respectively, independently performed IBR. Fusion time and registration error were recorded. Fusion success was defined both globally (success by all or at least one operator) and individually (registration error < 10 mm). Clinical and anatomical factors were assessed. Predictors of failure were identified using multivariable logistic regression with Firth's correction.

Results: IBR was successful in all three operators in 86.9% of cases and by at least one operator in 96.4%. The most experienced operator achieved significantly shorter fusion times (median: 15.0 s) and smaller fusion errors (median: 6.0 mm) compared to the less experienced operators. Operator-specific success rates defined as registration error < 10 mm were 45.2%, 60.7%, and 79.8%, respectively (p < 0.001). Subcutaneous tissue depth was the only independent predictor of fusion failure in both multivariable models (OR = 1.13 for all failed, p = 0.033; OR = 0.88 for partial success, p = 0.012). Other clinical factors were not statistically significant.

Conclusion: IBR is a highly feasible method that reduces operator dependency in liver fusion imaging compared to conventional methods, though registration accuracy still varies with operator experience.

Clinical impact: IBR enables consistent and simplified fusion imaging regardless of operator experience. Its broad applicability may support safer and more efficient ultrasound-guided interventions, especially in resource-limited or time-sensitive settings.