Artikel
A fully automated registration for 3D-Navigation
Suche in Medline nach
Autoren
Veröffentlicht: | 8. August 2007 |
---|
Gliederung
Text
Introduction: Intraoperative 3D-navigation [Ref. 1] is accompanied by the potential sources of error, arising from the manual registration. Many approaches require the fixation of the patient, which increases the invasiveness of the intervention. We suggest, that the surface of the occipital bone is acquired intraoperatively with A-mode ultrasound and used for registration, where no human interaction is necessary.
Methods: The transducer was positioned at the lambda-fissure of the occipital bone with stepper motors and sub-millimetric accuracy (Fig. 1 [Fig. 1]). The motion was controlled within LabView. The signal was further processed in MatLab, to generate a 3D-surface. A human skull was scanned with the laboratory setup. The acquired surface was registered with the iterative closest point (ICP) algorithm [Ref. 2] to a segmentation of the pre-operative dataset.
Results: The surface of the lambda-fissure was acquired in a rectangular area of 40 x 50 voxels with the spatial resolution of 1 mm (Fig. 2 [Fig. 2]). It contained detail from the specific relief of the selected anatomic area, which was of advantage for the registration. Some noise was present in the surface, where its inclination prevented the successful receiving of the echo-signal, e.g. at points of sutures on the skull. The preoperative dataset was segmented automatically with the approach described in [Ref. 3]. The two surfaces were registered with the ITK-implementation of the ICP algorithm. The registration took 7 h on a 2.66 GHz, 256 MB RAM PC, with 47 000 spatial points in the fixed dataset and 2 000 in the moving one. The accuracy of the registration was evaluated with the pointwise distance between corresponding points. The pointwise distance was displayed in a color map (Fig. 3 [Fig. 3]), which allowed the accuracy to be visually evaluated. The colors in the scale correspond to distances of ≤ 0.25 mm; ≤ 0.50 mm; ≤ 0.75 mm; ≤ 1.0 mm and > 1.0 mm, respectively.
Conclusion: The laboratory experiments were a proof-of-concept for the method. It is apt to improve the clinical accuracy of 3D-navigation, and to completely avoid patient fixation. Its speed is to be optimized in the next step of the study, where it will be applied on a cadaver.
Acknowledgement: This study was partially supported by the Jubilee Funds of the Austrian National Bank under project 9318.
References
- 1.
- Gunkel AR, Freysinger W, Thumfart WF. Computerunterstützte 3D-Navigationssysteme: Überblick und Standortbestimmung. HNO. 2000;48:75-90
- 2.
- Besl PJ, Mckay ND. A Method for Registration of 3-D Shapes. IEEE transactions on pattern analysis and machine intelligence. 1992;14(2):239-56.
- 3.
- Diakov GM, Freysinger W. Accuracy evaluation of initialization-free registration for 3D-navigation. Int J CARS. 2007 (accepted for publication).