| Background:Anterior cruciate ligament（ACL）of the knee joint is an important anatomical structure connecting the tibia and femur,essential in limiting the excessive forward movement of the tibia and maintaining the stability of the knee joint,together with other ligaments.The ACL rupture can lead to evident knee instability,seriously affecting knee joint function.If it is not treated in time,it is not only accessible to cause repeated knee sprains but also may cause secondary injuries to important knee joint structures such as articular cartilage and meniscus,leading to premature degeneration of knee joint and osteoarthropathy.Based on the existing research,arthroscopic ACL reconstruction is the mainstream surgical method for ACL injury.The purpose of this operation is to reconstruct the ACL,restore the stability of the knee joint to obtain better sports performance,and,at the same time,protect other structures of the knee joint from injury（especially the meniscus and articular cartilage）.Establishing graft bone tunnels is the critical step of ACL reconstruction.The research shows that the main factors affecting the prognosis of ACL reconstruction include the location of the graft tunnel,the graft’s choice,the graft,the fixation method of the graft,and the postoperative rehabilitation scheme.Among them,the poor location of the tunnel is the main reason for the failure of ACL reconstruction.The poor location of the tunnel causes about 70%-80%of the ACL reconstruction failures.Therefore,how to locate and orient the graft tunnel efficiently and accurately is the focus of arthroscopic ACL reconstruction.In recent years,the research on technical systems such as virtual reality（VR）and mixed reality（MR）has been deepened and gradually applied to many aspects of human production and life to bring great convenience.VR is more commonly used in surgical skills training,preoperative preparation,and doctor-patient communication in medicine.It allows medical students and surgeons to practice in a virtual controlled environment to improve surgical skills,deepen anatomical understanding,and avoid the risk of making serious mistakes.It can also significantly improve the information inequality between doctors and patients and effectively improve communication between doctors and patients.Whereas,MR is more likely to be used to guide surgeons during operation,which has the potential advantages of providing clear anatomical relationship diagrams in real-time,effectively improving the accuracy of operation,reducing the difficulty of operation,reducing the exposure risk of patients and doctors during operation and improving clinical results.Therefore,applying MR in orthopedic surgery has excellent development prospects.However,no related research report exists on applying MR in ACL reconstruction.Image matching algorithm is essential to improving the accuracy of image registration in the medical application of mixed reality technology,and the development of image matching algorithm is the main direction of future research.Part one:Clinical application of anterior cruciate ligament reconstruction assisted by mixed reality technology.Object:To evaluate the accuracy of tunnel position and direction and early clinical effect in ACL reconstruction assisted by mixed reality technology.Methods:A retrospective case-control study was conducted.After applying the selection criteria,44 patients with ACL injuries admitted by the Department of Orthopedics of the Army Medical Center of PLA of China of the Army Medical University from June 2020 to March 2022 were included,including 21 patients in the mixed reality technology-assisted surgery group（MR group）and 23 patients in the conventional surgery group.All patients underwent arthroscopic single-bundle ACL reconstruction.After the operation,the parameters related to the graft bone tunnels were compared by CT imaging and three-dimensional reconstruction.The early postoperative clinical effect was evaluated by instructing patients to fill in the Lysholm functional scales and the International Knee Documentation Committee（IKDC）scales.Results:There was no statistically significant difference in the projection angles of the actual postoperative tunnel and the preoperative simulated tunnel guide pin in the MR group on each axial plane（all P>0.05）.Compared to the conventional group,the MR group had the femoral tunnel exit location closer to the apex of the lateral femoral condyle in the proximal-distal axis（P<0.05）and a minor graft bending angle（P<0.05）.Moreover,the scatter plot of the distribution of femoral tunnel center position showed that the points in the MR group were less discrete and closer to the ideal location of the preoperative design than in the conventional group.The mean operative time（P=0.304）and the location of the tibial tunnel exit center at the tibial plateau(PML=0.602,PAP=0.473)between the two groups were not statistically significant.All patients were followed up for over 12 months,with an average of 17.91±4.49 months.The differences in the Lysholm functional score and the IKDC score of the affected limb between the two groups at 3,6,and 12 months postoperatively were not statistically significant（all P>0.05）,but both were significantly improved compared to the preoperative period（all P<0.001）.Conclusion:ACL reconstruction with the aid of MR technology allows for accurate positioning and orientation of the femoral tunnel during surgery,effectively improving the accuracy of femoral tunnel reconstruction,making it one option for individualized ACL reconstruction.According to our search,this study is the first literature report on mixed reality technology-assisted anterior cruciate ligament reconstruction.Part two:Preliminary exploration of medical image registration algorithm based on bounded generalized Gaussian mixture model.Object:To explore new medical image registration methods and provide theoretical basis for the next step of mixed reality technology combined with navigation to achieve automatic intraoperative image registration and improve image registration accuracy.Methods:A bounded generalized Gaussian mixture model（BGGMM）was used to approximate the joint intensity of the source medical image.The formulation of the mixture model is based on a maximum likelihood framework and is solved by an expectation-maximization algorithm.The largest musculoskeletal imaging（MURA）（https://stanfordmlgroup.github.io/competitions/mura/）database open source from Stanford University and the Anatomic Tracings of Lesions after Stroke（Atlas）（http://fcon＿1000.projects.nitrc.org/indi/retro/atlas.html）open source from the University of Southern California were selected to test the registration performance of this registration method and traditional registration methods.The pixel displacement（PAD）is used as the registration error to objectively measure the registration performance of different methods.Results:A medical image registration method based on bounded generalized Gaussian mixture model（BGGMM）was proposed,using two open source datasets,MURA and Atlas,for simulation and training.Compared with traditional registration methods such as mutual information method,enhanced correlation coefficient method,and ensemble registration method,the image registration performance of this method has significantly improved.Compared with other methods,the average PAD values of the MURA dataset decreased by88.15%,97.21%,and 96.52%,respectively.Compared with other methods,the average PAD values of the Atlas dataset decreased by 5.12%,93.68%,and 92.55%,respectively.Conclusion:The proposed BGGMM registration method significantly improves the registration performance of medical images,and its effect is more obvious when processing source images with more interference information and larger offset.This provides a theoretical basis for further developing three-dimensional registration models based on this method in the future,combining with navigation systems,to achieve automatic registration of intraoperative mixed reality medical image models and real entities. |
PDF Full Text Request