A Simple Structural Magnetic Resonance Imaging (MRI) Method for 3D Mapping between Head Skin Tattoos and Brain Landmarks Mulugeta Semework State University of New York, Downstate Medical Center, Brooklyn, USA
Abstract— A successful brain surgery requires pre-surgery localization of various brain areas. An accurate craniotomy, which gives perfect access to such brain area of interest is needed and is dependent on mathematically establishing the relationship between head landmarks and structural magnetic resonance image (MRI) scans. While typical stereotactic procedures rely upon external cranial landmarks and standardized atlases for localization of subcortical neural regions, visualization of the internal morphology of the brain in vivo can be achieved by MRI. Our lab, when possible, also uses MRIs on our post-surgical monkeys to get precise information on the exact placement of implanted microwires. Our stereotactic instrument and head-posts on the monkeys are compatible with a magnetic resonance unit and we have developed a model to analyze the magnetic resonance imaging results and calculate 3D mappings between external landmarks, skin tattoos on the primates’ heads and brain areas of interest. This allowed us to overcome the limitations, inaccuracies, and cost prohibitions of the traditional stereotactic methods and helped us make reliable localization of subcortical targets in the monkey brain. Keywords— Structural MRI, Structural MRI image analysis, brain structure mapping, Stereotactic method, neurosurgery.
I. INTRODUCTION With increased ease, power and accessibility of magnetic resonance imaging (MRI), especially the high resolution of T1-weighted structural MRI [1] there has been a growing interest in using this technology to study brain structure, function, development, and pathologies [2]. In what is now a very common practice of implanting primates with microelectrode arrays (MEAs) for various research goals, there is a common problem that arises from the inherent variability of brain structures and skull anatomy between different subjects. Since MRI is a non-invasive method that capitalizes on the complex mosaic across the cortical sheet [3], it is possible to solve much of this problem by taking individual MRIs pre-surgery and be able to map structures of interest and use established coordinates during the surgery. There is variability between individuals in pattern of brain area folding, shape and size of cortical areas and relative locations
[3]. We and others observe individual differences in brain anatomy even though the overall organization and relative location with respect to each other stays the same. It is very common to find errors in subjective guesses of location of a brain structure just from skull topography alone. Moreover, as standardized atlases are generally used for localization of subcortical neural regions [4] a problem arises from such poorly informed assumption of the location of underlying brain structures and it is not uncommon to make a misplaced craniotomy. There is thus a need for a method to make a reconstruction of the areas of interest and describing the relationships within a reasonably acceptable mathematical error. This short paper discusses a recently developed new method for expressing relationships between surface markers, such as tattoos on head skin and underlying major brain structures.
II. METHODS A. MRI Procedure Monkeys (Macaca radiata) are anesthetized in their home cage with Ketamine (10-20 mg/kg) injection intramuscularly. We found that most of our monkeys remain under for the duration of the scanning by just this drug alone. For the MRI procedure, the anesthetized monkeys are transported to the SUNY Downstate Medical Center (DMC, also known as “University Hospital of Brooklyn”) Department of Radiology. In the facility, after their head is stabilized with earbars, the monkeys are placed into the scanner chamber, and their heads fitted inside a 16-in. head coil. Monkeys remain anesthetized during the MRI procedure, if needed, with a supplemental injection of Ketamine. Since the procedure takes only 30-50 minutes, the first anesthesia injections are generally effective in maintaining stillness inside the machine. MRIs of brains are acquired on a Magnetom Symphony Maestro Class Scanner, and the following parameters are for a typical scan and can vary between monkeys and scans (2). T1-weighted 3D MPRAGE MR images are acquired through the entire brain using a TR = 1,500 ms and
K.E. Herold, W.E. Bentley, and J. Vossoughi (Eds.): SBEC 2010, IFMBE Proceedings 32, pp. 493–495, 2010. www.springerlink.com
494
M. Semework
TE = 3.04 ms with no echo-train. Scan acquisition time will approximately be 10 min. For each monkey the number of signals were averaged to be three. Slices will be obtained as 0.5-mm thick contiguous sections with a matrix size of 256 × 256 and a field of view of 128 mm × 128 mm, resulting in a final voxel size of 0.5 mm × 0.5 mm × 0.5 mm. After being transported back to their cages, monkeys are allowed to completely recover from the effects of the anesthesia. All procedures are done per SUNY DMC’s Animal research regulations and the Joseph T. Francis lab protocols.
C. Image Processing To help in the ease and perfection of the next procedures, the scans are first enhanced, region of interest (ROI) cropped, and then appropriately edged, a subjective process that depends also on the quality of the scans and the resulting binary images. Depending on several factors, the threshold for edging can be manually set.
B. Code and User Interface The image metadata from a DICOM file series is made into a structure, a 3D array of the images is generated, and previewed using a Matlab tool (5). All of the analysis is done by a novel Matlab function that imports the DICOM images and does the preprocessing and detailed calculations. Currently, there is a tested command line version of the code. The most recent one is designed for a user with no Matlab programming skills and is still being optimized. In short, it has a Graphic User Interface (GUI) used as a front panel to make option selections, such as importing slides, color maps, edging methods, etc. that makes the analysis easy and user friendly (Fig.1)
Fig. 2 Sobel-edged and selected brain, skull and marker surfaces As shown in Fig. 2 (which shows already highlighted targets, in color), the edged scan is now ready to be pointand-clicked, or, in case of the new code (GUI shown in Fig. 1), a ROI is first manually selected, and an automated tracking of this object throughout all the scans follows. There is the option of deciding how many connected points to consider in the analysis, colors to use, etc. in graphical displaying of the process. D. Transformations The user has the option of selecting skull or brain landmarks to be used for this purpose. The outside markers we use, skin tattoos, are identified in the MRI scans from vitamin E tablets we affix to the skin during the scanning (yellow ellipses on Fig. 3).
Fig. 1 Graphic User Interface for new MRI mapping method (a “Canny” edged brain scan shown)
IFMBE Proceedings Vol. 32
A Simple Structural Magnetic Resonance Imaging (MRI) Method for 3D Mapping
Once the outside and inside points are set, they are tracked through all or few selected scans that are known to contain them. The absolute distance between any two markers (mDist) in the scans 3-D space is calculated as follows:
mDist{x, y, z} =
(
the skin landmarks which made it difficult to do sterile procedures during surgery. These markers ended up hiding deep in the stereotactic apparatus and therefore only two locations had to be used, still with no error in location the target.
( | (x 1 - x 2 ) |2 +
| (y1 - y 2 ) |2 ) * pw + (| (s1 - s 2 ) | *th)2 ) Where x and y are the locations of the given markers (or ROI and marker), “th” is scan thickness in mm and “pw” is pixel width. Following this same convention, the new automated approach computes the normalized 2-D crosscorrelation of the target matrix template and the cropped scan and finds the maximum coefficient and gives mDist in same 3-D space.
495
IV. CONCLUSIONS For the sole purpose of making a craniotomy above the location of the brain structure we are interested in, we found this method to be very useful and dependable. It eliminated the need to widen craniotomies or make new ones to correct errors. The current code refinement is expected to make this method available for wide use and help improve structural mapping success. A broader application of this method is envisioned for structures that are hidden from view and reach or are too topologically convoluted.
III. RESULTS AND DISCUSSION
ACKNOWLEDGMENT
Out of the need to make a simple craniotomy that lands on the brain structures that we are interested in implanting multielectrode arrays, the above method was developed. It is designed to map physical markers, tattoos on the monkeys’ heads, and the underlying major brain landmarks, such as the central sulcus. As shown in Fig. 3 (left panel), a pencil sketch on the skull of the predicted location of the central sulcus (diagonal trace) corresponded well with the actual finding after the craniotomy.
I thank everyone in the Joseph T. Francis lab at SUNY DMC for their unreserved support in taking the MRI scans and comments in developing this method. Westley Hayes, thank you so much for your professional and speedy manuscript editing.
REFERENCES 1. Saad Z, Glen D, Chen G, et al. (2009) A new method for improving functional-to-structural MRI alignment using local Pearson correlation. NeuroImage 44, 839–848 2. Smith S., Jenkinson, M, Woolrich, M, et al. (2004) Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage. 23 (2004) S208–S219 3. van Essen, D, Drury, H, Joshi, S, and Miller, M (1998) Functional and structural mapping of human cerebral cortex: Solutions are in the surfaces. Proc. Natl. Acad. Sci. USA. Vol. 95, pp. 788–795 4. Saunders, R, Aigner, T & Frank, J (1990). Magnetic resonance imaging of the rhesus monkey brain: use for stereotactic neurosurgery. Exp. Brain Res. 81, 443–446 5. Balkay, L (2005) DICOM Reader at http://www.mathworks.com/matlabcentral/fileexchange/ 7926-dicomdir-reader
Fig. 3 Craniotomy sketch and matching central sulcus As a proof of individual differences, we tried using the coordinate system and marker distances that were generated from one monkey on another one and the craniotomy was off by at least half a centimeter. One challenge we faced was the lower placement on the side of the face of some of
Author: Institute: Street: City: Country: Email:
IFMBE Proceedings Vol. 32
Mulugeta Semework State University of New York, Downstate Medical Center 450 Clarkson Ave, box 31 Brooklyn USA
[email protected]
