Abstract / Summary
An asymmetric hybrid registration frame (RF) integrates radiopaque and optical markers to automate computed tomography (CT) registration and coordinate hand-over to a dynamic reference frame (DRF). We compared automatic and manual fiducial selection using three fixed CT datasets of a rigid sawbone phantom. Ten operators performed five manual registrations per CT, alongside one automatic benchmark per CT. Both methods used the same five registration fiducials, rigid registration procedure, and three evaluation centers excluded from transform estimation. Target registration error (TRE) was the root mean square distance between transformed design centers and CT-detected centers, without probe contact. Across equally weighted CT datasets, mean TRE was 0.538 mm for automatic and 1.565 mm for manual registration. The manual-minus-automatic difference was 1.027 mm (95% operator-bootstrap confidence interval, 0.951–1.148 mm). Mean registration processing times were 2.817 and 22.589 s, respectively. Under these common evaluation conditions, automatic fiducial detection reduced on-frame center discrepancy and registration time. These findings support the integrated registration and handover workflow as a basis for further validation at independent anatomical targets.