Kinect Camera Based Navigation For Percutaneous Liver Needle Intervention

Percutaneous biopsies and ablations are the daily clinical practices for liver tumor diagnosis and treatment,while accurate needle insertion is a major difficulty to perform percutaneous procedures.An accurate needle insertion procedure can be achieved through the surgical navigation system,which can map the needle into preoperative CT(computed tomography)image space to guide a surgeon to accomplish the puncture procedure with the improved treatment outcome.The abdominal puncture navigation has been developed for more than a decade,but it is barely applied to clinical practice,because the techniques utilized for implementing the system(like registration,respiratory motion compensation,etc.)are limited.Recently,the real-time surface imaging techniques have been gradually applied into soft-tissue navigation,and exhibit good potentials of upgrading the conventional methods.Considering the advantages of new imaging modalities,a Kinect camera was integrated into an abdominal puncture navigation system to investigate novel solutions on registration and respiratory motion compensation.The main contributions of this thesis are listed as follows:(1)A needle insertion navigation system with markerless registration was developed based on the second generation of Kinect camera.In this work,the Kinect camera was integrated into a conventional needle insertion navigation system successfully,and a two-dimensional shape image based abdominal surface matching algorithm was proposed to implement the real-time markerless registration.To perform the comparison between different registration approaches,an automatical marker based registration was proposed.Furthermore,two generations of Kinect cameras were integrated into the home made system to investigate the differences between two Kinects on guiding needle insertions.With the phantom and animal experimental evaluation,the results indicate that the system accuracy can meet the clinical requirements and is improved compared with that of previous studies considerably,the markerless registration obtains higher navigation accuracy than the marker based registration,the second generation Kinect outperforms the first generation in the accuracy of guiding needle insertions.(2)In vivo comparison between three different needle insertion guidance methods was conducted.To validate the feasibility of clinical application of the proposed navigation system,needle insertions were performed under guidance of the proposed system(Kinect-Optical navigation),the conventional CT image guidance approach,and a commercial navigation system with five healthy beagles.The experimental results show that the puncture accuracy of Kinect-Optical system is higher than that of the conventional CT-guidance and is comparable to the commercial system.The time cost of entire puncture procedure for the Kinect-Optical navigation is similar with the other two guidance approaches.(3)A motion modeling framework combining preoperative statistical respiratory motion model with intraoperative abdominal surface matching was proposed.For compensating respiratory motion of abdominal organs,a PCA(Principal Component Analysis)based statistical motion model was combined with a metric of abdominal surface similarity to build a respiratory motion model.To improve the motion estimation accuracy,a metric based on Bayesian estimation and a preoperative statistical motion model based on Kernel PCA was integrated into the proposed motion model,respectively.The initial experimental results show that each of the three motion models obtains a similar motion estimation error with state-of-the-art studies,and has a good potential of dealing with the problem of irregular breathe adaption.