The palpation of the controlateral iliac spinae remains a major hurdle to the success of navigation in lateral position. Several studies are seeking for alternative landmarks to compute the anterior pelvic plane (APP). Up to now, none of those methods have been used in clinical routine. Ultrasound navigation offers a great potential to identify new bony landmarks. The tubercles of the lower lumbar spine and the symphysis can easily be imaged. Those points define a sagittal plane, that can be used as a symmetry plane to compute a virtual controlateral spinae from the acquired colateral spinae. A virtual pelvic plane can then be computed. The objective of this study was to check the accuracy and reproducibility of this virtual anterior pelvic plane. 6 hips (3 left, 3 right) from 4 cadavers (mean BMI 22,6; range 19,5–26,7) embalmed with glycerol and alcohol were used for this study. All anatomic landmarks were acquired with the OrthoPilot® Ultrasound navigation system. One experienced surgeon acquired the reference APP with the cadavers lying supine. The cadavers were then placed in lateral position. Two experienced surgeons acquired 6 times following landmarks: 3 lower lumbar tubercles, 3 sacral tubercles (see Figure 1), the posterior spines, the symphysis and the colateral iliac spine. Several sagittal planes were computed using all points (least square plane) and all possible combinations between one symphysis point, one lower lumbar tubercle point (L5, L4 or L3), and one sacral tubercle point (S2 or S1). The angular error of the resulting virtual APP to the reference APP was computed. For each cadaver, an error map was computed to visualize the error of the virtual APP with respect to the height of the used sacral and lumbar tubercles along the spine. The reference APP was acquired with a good reproducibility: the deviation between each acquisition to the mean of all acquisitions was smaller than 1° (except for cadaver 2 right side, the deviation reached 2 ° in the frontal plane). As some sacral and lumbar points were mixed during the acquisition, the line joining the posterior spines was used to separate the sacral from the lumbar points. The mean errors and standard deviations were comparable between operators. The least square plane computed with all points strongly depended on the cadaver positioning : for the same cadaver, the mean error reached 0°on the left side and 8° on the right side. More constant results were obtained by using a combination of 3 points. 5 outliers were identified and removed as they clearly corresponded to erroneous acquisitions on bad quality images. After having removed those outliers, the mean error ranged between 2° and 5° and the standard deviation between 1° and 3°. The best combination of points was a point on the symphysis, the lowest sacral tubercle (S2) and the lowest lumbar tubercle (L5). This study shows that the symphysis, the lower lumbar and sacral tubercles can be used to define a sagittal plane and thereby define a virtual anterior pelvic plane. Outliers should be suppressed by taking special care to the image quality and by adding a guided ultrasound functionality: visualizing the resulting sagittal plane on the ultrasound picture would enable the surgeon to easily control the accuracy of his acquired plane. The next steps consist in checking the feasibility in a clinical set-up.
