|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.|
Our data demonstrate the feasibility of introducing a new automated registration protocol into the workflow of transcranial US fusion imaging. The automatic registration algorithm worked in all examined subjects with excellent positional correspondence in major intracranial color Doppler-accessible vessel segments, i.e. the A1-ACA, M1-MCA, M2-MCA and ICA-Siphon.
Time needed for the registration procedure was, despite the use of a study workflow only, not longer than the time needed for the conventionally used external or internal marker matching procedure. A detailed analysis was not performed because of the feasibility character of the current study and would only be sensible after optimized workflow integration into the matching procedure.
By achieving the two above aims, the Virtual Navigator Automatic Registration Algorithm has the potential to become a valuable solution for routine insonation and follow-up sessions performed even by different operators with different skills. An expert, but especially a non-expert examiner, who is not trained in manual matching procedures, will be able to perform precise tuning on the small vessels of the described region. Indeed, the residual error after the automatic registration was zero in most of the cases, otherwise of 1 mm and it was of 3 mm in one case only. As a reference, the average dimensions of the considered vessels were between 0.5 and 2.5 mm for A1-ACA, M1-MCA and M2-MCA with the ICA-Syphon in average had a diameter of 2.5mm at the insonation level considered. Although not formally analyzed and addressed in the current pilot study, this approach not only increases the confidence of image alignment achieved by the technique but also has the potential to improve repeatability of matching results for follow-up investigations.
Compared to previously described registration procedures, which consisted in initial fiducial markers registration, with the registration procedure described in this work , there is no need to point external landmarks on the patient’s face, and neither to move the receiver from the Registration Pen to the Probe.
Choosing a transcranial insonation paradigm, the proposed Automatic registration algorithm showed to be able to work in rather difficult insonation conditions. Transcranial US insonation frequencies are low, resulting in a low correspondingly spatial resolution. The success in these conditions suggests an even easier application if used for insonation of other – e.g. extracranial – organ systems, provided that vascular tree is visible within the field of interest. Also, the chosen automated segmentation of MR derived vessels should be transferrable to other organ systems and even other image modalities, e.g. a vascular tree derived from a CT-angiographic data set.
Currently, the 2-step approach of initial manual pre-registration is required to reduce the demand on computational power for the second, automated part of the registration process. Future optimization might be able to proceed even without this extra step within the workflow.
A number of potential pitfalls related to the proposed algorithm has to be considered:
- First, visualization of one or more vessel bifurcations is mandatory to reach an optimal matching result. Consequently, the presented solution will not work in typical muscle, skeletal, prostate, gynecological US due to the lack of clear vascularization landmarks. Currently, the automated vessel recognition and segmentation only work on white vessel signals, so a TOF MRI sequence or a post-contrast T1-MRI sequence are required.
- Second, vessel size between the two datasets varies. MR-derived vessels are small - US Color Doppler derived vessel are larger, depending on system settings like persistency, smoothing level, possible movements and general acquisition errors of the US system. This discrepancy can lead to an incorrect alignment of the US and MRI vessel as the mathematical overlapping region between the vessel tree has more than one “best fit”. Although not formally analyzed, this phenomenon is most likely the reason for the observed discrepancies between our analyzed data sets.
- Third, the matching outcome was satisfactory for vessels analyzed in the area ipsilateral to the side of the registration procedure. However, if contralateral vessels were measured, the observed error increased up to more than 5mm (data not shown). Possible reasons of this mismatch could be:
- Persistence of the Color Doppler signal – the longer the distance the US has to travel, the higher the persistence effect and subsequent time-space delay, the larger the mismatch.
- Geometric deformation of the US beam caused by beam configuration of the Phased array probe but also caused by bone window induced beam distortion increases with rising insonation depth.
- Depth dependent loss of Color Doppler signal intensity and spatial resolution with subsequent depth dependent varying quality of automated vessel tree calculation. This phenomenon might potentially be overcome by the use of intravenously applicable US Contrast media (provided that there are no contraindications).
- Technical accuracy - the electromagnetic antenna technology is guaranteed by the manufactures for 1.5mm maximum accuracy only.
- Inhomogeneity of the electromagnetic field – application of the technique in a typical hospital setting implies the nearby presence of metallic structures, other electric instruments or wires which might all result in field distortion. This phenomenon will even to some degree be caused by the insonated object itself, lying in the field and by the moving insonation probe too.
Looking at small structures like the circle of Willis, where the analyzed vessel structures have original diameters of 0,5-2mm, the above minor effects might sum up to the observed errors.
In conclusion, automatic registration seems to be a feasible, fast and precise method to obtain fusion imaging between US and MRI images in Neurosonology application. This method holds the potential of offering a faster and easier way to obtain a precise registration. Further studies are necessary to confirm our preliminary results.
Thematically related posters
ECR 2015 / C-0264
Imaging of the anterior and central skull base as a guide for endoscopic skull surgery