Abstract
Introduction
Bony deformities in the hip that cause femoroacetabular impingement (FAI) can be resected in order to delay the onset of osteoarthritis and improve hip range of motion. However, achieving accurate osteoplasty arthroscopically is challenging because the narrow hip joint capsule limits field of view. Recently, image-based navigation using a preoperative plan has been shown to improve the accuracy of femoral bone surfaces following arthroscopic osteoplasty for FAI. The current standard for intraoperative monitoring, 3D x-ray fluoroscopy, is accurate at the initial registration step to within 0.8±0.5mm but involves radiation. Intraoperative 3D ultrasound (US) is a promising radiation-free alternative for providing real-time visual feedback during FAI osteoplasty. The objective was to determine if intraoperative 3D US of the femoral head/neck region can be registered to a CT-based preoperative plan with comparable accuracy to fluoroscopic navigation in order to visualise progress during arthroscopic FAI osteoplasty.
Methods
The experiment used a plastic femur model that had a cam deformity on the femoral head/neck. Thirty metal fiducial markers were placed on the US-accessible anterior and lateral surfaces of the femur. A CT image was acquired and reconstructed, then used to develop a preoperative plan for resection of the cam deformity. Twenty-two sets of 3D US data were then gathered from the phantom using a clinical ultrasound machine and 3D transducer while the phantom was submerged in water. US surfaces from the anterior/lateral regions of the femur were extracted using a recently proposed image processing algorithm. Fiducials in the US volume were manually registered to corresponding CT fiducials to provide a reference standard registration. The reference standard fiducial registration error (FRE) was measured as the average distance between corresponding fiducials. After fiducial-based registration, each US surface was randomly misaligned and re-registered using a coherent point-drift algorithm. The resulting surface registration error (SRE) was measured using average distance between US and CT surfaces. Finally, a plastic model of the preoperative cam deformity resection plan was 3D-printed to represent the postoperative femur. Five US scans were acquired of the postoperative model near the femoral head/neck. Each US scan was initialised for 20 trials using three reference points, and then registered using coherent point drift. Surgical outcome accuracy was reported using final surface registration error (fSRE).
Results
The reference standard FRE was 0.41±0.19mm. The distance between surfaces following misalignment and re-registration for all 2200 automated registration trials was similarly small (SRE = 0.31±0.04mm) and well below the required clinical limit. Lastly, the postoperative model was accurately registered to corresponding US scans (fSRE = 0.58±0.07mm). Qualitative visualisation showed good surface matching following US to CT registration.
Conclusion
Initial registration between intraoperative 3D US and preoperative CT is critical for accurate visualisation of surgical progress during FAI osteoplasty. Given spatial initialisation, the achievable registration accuracy of 3D US to CT is 0.31±0.04mm (SRE) which is well within the fluoroscopy standard, 0.8±0.5mm. The results suggest strong potential for ultrasound to guide computer-assisted arthroscopic FAI osteoplasty.
