Artikel
Registration of the lumbar spine based on 3D ultrasound and spiral-CT data
Ultraschall- und Spiral-CT - basierte Registrierung der Lendenwirbelsäule
Suche in Medline nach
Autoren
Veröffentlicht: | 4. Mai 2005 |
---|
Gliederung
Text
Objective
Intraoperative navigation in spinal surgery is usually based on flouroscopy or uses intraoperative CT for referenciation. These techniques are both time consuming and due to radiation need special requirement in the operating theatre. The key to the implementation of ultrasound in this process is the successful data work-up to allow referenciation. The aim of this study was to find out wether a referenciation is possible and what alogorithm would be the best for this purpose.
Methods
In a first step a model of the lumbar spine was used to compare 6 different algorithms for referenciation. Optimal results were obtained using the CMA (Covariance Matrix Adaptation) evolutionary strategy for optimization. In a second step 3 patient data sets were evaluated. Data fusion was done using 3D-ultrasound of the lumbar spine and a spiral-CT of that region. Each vertebra was registered 1000 times with different starting positions. These starting positions where chosen relative to the optimal position with a translational deviation between 0 and 15 mm and a rotational deviation between 0 and 11 degrees. Results with a lower distance than 1 mm to the optimal position were defined as correct registration The registration algorithm is based on a surface-volume registration. The bone surface points were extracted semi automatically from the preoperative CT datasets for each vertebra. Only points where taken, which could also be visualised by ultrasound An enhancement of bony surface structures is done in the ultrasound data. The metric for the registration is the sum of the ultrasound gray values covered by the surface points.
Results
Using the patient data 9 different vertebras were referenciated according to the protocol. Regardless of the changing starting points CMA allowed a correct registration in 98.1% of the trials. 96.5% of the results had a distance below 0.1 mm to the bone surface. The mean distance of the correct trials was 0.0141 mm with a standard deviation of 0.0777.
Conclusions
Referenciation for spinal navigation based on 3D ultrasound of the lumbar spine is possible using the CMA evolutionary strategy for optimization. Further testing should allow a preoperative patient selection in regard to necessary image quality of the ultrasound data sets. Furthermore, intraoperative application with virtuell insertion of pedicle screws should allow a refinement of the hard- and software.