|ECR 2015 / C-0828|
|Virtual Navigator Automatic Registration Technology in Transcranial Application|
|This poster is published under an open license. Please read the disclaimer for further details.|
Methods and materials
Six (6) subjects (5 males, 1 female; age range: 36-86, mean age: 54 years) were included into the study. All had a patent transtemporal bone window for transcranial insonation and routine US yielded normal findings for all intracranial accessible basal cerebral arteries. In all individuals, pre-registered MRI images, obtained for reasons not related to the study, were available. The study was performed in the Ultrasound Lab of the Dept. of Neurology (Charité- Universitätsmedizin Berlin, Germany) using a commercially available US system (MyLabTwice, Esaote S.p.A. Italy), equipped with Virtual Navigator (VN) option  and additional local installation of the automatic registration algorithm.
Methods of examination
Brain MRI was acquired from all subjects using a 1.5 Tesla scanner (Siemens Magnetom Avanto, Erlangen, Germany), equipped with a 12-channel head coil. The following sequences were available and used for registration: 1) 3D T1-weighted MP-RAGE for anatomical comparison; 2) 3D-multislab time of flight magnetic resonance angiography with a cranial saturation band, in order to suppress venous and highlight the arterial signal. The parameters of the two sequences were: 1) repetition time (TR)=2650 ms, echo time (TE)=28/113 ms; echo train length=5; flip angle=150°, 50 axial slices with a matrix size=256x256, interpolated to 512x512, field of view (FOV)=250x250 mm2, slice thickness 2.5 mm; 2) TR=1900 ms, TE=3.37 ms, TI=1100 ms, flip angle=15°, 176 contiguous, axial slices with voxel size=1x1x1 mm, matrix size=192x256, FOV=192x256 mm and slab thickness=176 mm.
Fusion imaging was carried out with the above US system using a Phased Array Probe (Operating Bandwidth: 1–4 MHz, PA240, Esaote) and a reusable tracking bracket with sensor mounted (639-039, CIVCO Medical Solutions, Kalona, Iowa, USA). MRI DICOM data sets were transferred to the US VN via DVD import or via network based import from the hospital PACS System (Centricity®, GE medical Systems) prior to the matching procedure.
The setup was with the patient lying in a supine body position with the head positioned within the electromagnetic field; source tip positioned to the right of the subject and pointed towards the head in order to achieve the highest homogeneity of the created field in the US scanning area. Two additional small receivers were used. One, attached to the patient forehead with a plaster strip (Figure 1) provided motion correction information, correcting any of the patient movements after completed registration procedure . The second was mounted on the US probe fixed by a support, providing position and orientation of the US probe in relation to the transmitter in the created 3D space. Before starting, an accuracy check of the electromagnetic field was performed: the same point coordinates were measured twice by a dedicated registration pen with the electromagnetic sensor mounted in two different spatial orientations. An accuracy of 0.2 cm or less was considered acceptable.
Color-coded US vessel images of the subject’s circle of Willis were obtained by transtemporal insonation through the right temporal bone window. Matching of US and the MRI data set was achieved in two workflow steps: manual pre-registration (one plane/one point) and secondly automatic registration of intracranial arteries. The details of both steps are explained in the following paragraphs.
1- Manual pre-registration (one plane - one point)
Similar axial planes – displaying the circle of Willis - were chosen on the US scan and the TOF-MRI dataset. One-point registration resulted in a first rough data set matching, i.e. moving the US images already yielded in real-time by the US probe with simultaneous navigation within the MRI data volume. Then one identical point well visible on US and MRI – the beginning of the M1-segment of the middle cerebral artery (MCA) - was marked on both image modalities. A subsequent one-plane registration resulted in an improved matching, however only adjusting the spatial error in the X, Y, and Z coordinates of this point (Figure 2). The procedure so far leaves a residual registration error with its magnitude depending on the accuracy of the manual matching point identification of the sonographer but this is a necessary precondition for the following automated registration algorithm .
2- 3D Automatic registration algorithm
The algorithm removing the residual shift of the two volumes is based on automated matching of the three-dimensional vessel tree visible in both modalities. For this, a prior automatic segmentation of the vessels was performed. In the TOF MRI sequence, the segmentation was started from the previously marked point as an initial seed (Fig. 3). Vessels were recognized based on their gray level similar to the one of the initial point, and their connection to it for at least a volume chosen by the operator (threshold of the segmentation). The three-dimensional US Color Doppler volume of the brain vascular tree (Fig. 4) was acquired and calculated with a similar mathematical approach with the data set acquired by a short 5 seconds sweep through the sonographic region of interest and the previously defined point as the initial seed. The following automated correction and registration step comprised a 3D-matching of both volumes using only the extracted, filtered and resampled vessels, through the downhill simplex algorithm, which maximizes the common part of the vessel trees.
For each subject, the accuracy of the registration was evaluated qualitatively and quantitatively. The former was achieved assessing the correspondence of the arteries visible on TOF MRI and on CD US (Fig. 5). The latter consisted on measuring the residual distance of the following anatomical points visible with the two modalities in the axial plane (Fig. 6): Anterior Cerebral Artery (A1-ACA), two points in the Middle Cerebral Artery (proximal point: M1-MCA and distal point: M2-MCA), and Internal Carotid Artery (ICA)-Siphon.
We tested if the residual error, measured for all the subjects in the four anatomical points, was statistically different from zero with the signed rank Wilcoxon test.
Thematically related posters
ECR 2015 / C-0264
Imaging of the anterior and central skull base as a guide for endoscopic skull surgery