PURPOSE To quantify the contribution of a computed tomography (CT) scan to navigation accuracy in computer-assisted surgery. METHODS Eighty-eight patients undergoing computer-assisted facial or skull-base surgery were fitted preoperatively with 4 to 12 markers, either attached to the skin (n = 20) or fixed in the osseous skull (micro-screws; n = 68). Low-dose high-resolution spiral CT was achieved with 25-cm field of view (FoV), 1-mm slice thickness, 2-mm table increment, 1-mm reconstruction interval, 140 kV, 40 mA, bony reconstruction algorithm, and 180 degrees reconstruction profile (effective slice thickness = 1.8 mm). During surgery, navigation accuracy was evaluated using two navigation systems. RESULTS Mean error was 0.66 mm for osseous markers and 1.58 mm for cutaneous markers. Both values are markedly smaller than the effective slice thickness of the scan protocol used. Radiation exposure of the patient for the entire examination never exceeded that necessary for one single 10-mm slice in a standard brain examination. Despite the reduced dose, landmarks and fiducials were precisely identified in all cases. CONCLUSIONS The CT-induced positioning error in the Z-axis is considerably reduced by overlapping raw data reconstruction. For 1-mm slices and a 25-cm FoV, the average scan-induced positioning error is about 0.3 mm. Spatial resolution is not affected by the low dose applied. For MRI-based navigation, a 1 mm3 voxel size is the best compromise between signal-to-noise ratio, spatial resolution and scan time.