We undertook a prospective, randomised study using a non-invasive transcranial Doppler device to evaluate cranial embolisation in computer-assisted navigated total knee arthroplasty (n = 14) and compared this with a standard conventional surgical technique using intramedullary alignment guides (n = 10). All patients were selected randomly without the knowledge of the patient, anaesthetists (before the onset of the procedure) and ward staff. The operations were performed by a single surgeon at one hospital using a uniform surgical approach, instrumentation, technique and release sequence. The only variable in the two groups of patients was the use of single tracker pins of the imageless navigation system in the tibia and femur of the navigated group and intramedullary femoral and tibial alignment jigs in the non-navigated group. Acetabular Doppler signals were obtained in 14 patients in the computer-assisted group and nine (90%) in the conventional group, in whom high-intensity signals were detected in seven computer-assisted patients (50%) and in all of the non-navigated patients. In the computer-assisted group no patient had more than two detectable emboli, with a mean of 0.64 (SD 0.74). In the non-navigated group the number of emboli ranged from one to 43 and six patients had more than two detectable emboli, with a mean of 10.7 (sd 13.5). The difference between the two groups was highly significant using the Wilcoxon non-parametric test (p = 0.0003).Our findings show that computer-assisted total knee arthroplasty, when compared with conventional jig-based surgery, significantly reduces systemic emboli as detected by transcranial Doppler ultrasonography.