Abstract
Introduction
A full 3D postoperative analysis, i.e. a quantitative comparison between planned and postoperative positions of bone(s) and implant(s) in 3D, is necessary for a thorough assessment of the outcome of the surgery, as well as to provide information that could be used to optimize similar procedures in the future. In this work, we present a method of postoperative analysis based on a pair of X-ray images only, which reaches a level of accuracy that is comparable with the results obtained with a 3D postoperative image.
Methods
The method consists in using 3D models of bones, segmented from 3D preoperative image (e.g. CT or MRI scans), and 3D models of implant, and aligning them independently to X-rays by matching contours manually drawn on the X-rays and projected contours. The result gives the relative postoperative position of bone and implant. The method was tested on a phantom consisting of commonly available femoral knee implant on a physical model of a femur (Sawbones®). Result was compared to the optical scan, considered as ground truth, of the implanted saw bone. Two studies were performed: inter-operator (six operators), and intra-operator (5 tests). In addition, the inter-operator study was repeated while asking all the operators to use the same pre-drawn contours. The results are presented by calculating the distance (anterior/posterior, proximal/distal, medial/lateral) between the centers of gravity, and the angles (varus/valgus, flexion/extension, external/internal rotations) of the implants from the X-ray based method and the ground truth.
Results were also compared with the relative position of bone and implant extracted from a 3D CT postoperative image. Saw bone and implant were first segmented from this image. In order to determine the position of the implant, despite the metal artefacts in the CT images, the 3D model of the implant was registered on the segmented implant.
All processing, including segmentation, registration of X-rays, and measurements, was performed using Mimics Innovation Suite 17.0 ®.
Results
Results are shown in Figure 1. The inter-operator study showed a maximum error of 0,55 mm and 1°. The intra-operator study, performed by an experienced operator, showed the importance of the learning curve, as all the measurement errors decreased (except for the varus/valgus that increased from −0,07° to 0.13°)) up to 0,13 mm and 0,23°. Same trend was showed by the results of the inter-operator based on same drawn contours, showing the importance of contour drawing on the results, and thus again the importance of the training. In comparison, results from postoperative CT showed similar order of magnitude with a maximum error of 0,65 mm and 0,61°.
Conclusion
The evaluation of our method gives preliminary good results, and was comparable with a postoperative analysis based on CT, while reducing the cost and the radiation dose of the analysis. A validation on cadaver and clinical data is planned to confirm our results.
