3D Digital MRM Atlas of Adult C57BL/6J Mouse Brain
- Sample Preparation
- MRI Imaging
- Individual Atlas Creation
- Minimum Deformation Atlas Creation
- Probabilistic Atlas Creation
- Group Deformation Map
- A “representative”, artifact-free MRI data
set was segmented manually into 20 regions of interest
(ROI) using Amira software.
- Other brain samples were automatically segmented using our pilot segmentation procedure and
- Pilot segmentations were manually refined using
- The structural volume and surface area were measured and stored as "Structure statistics" in the database.
Sample Preparation (In vitro)
Adult inbred C57BL/6J mice (12-14-week old males; weight range 25 -30 g; The Jackson Laboratory, Bar Harbor, ME) were anesthetized with an intraperitoneal injection of pentobarbital sodium 0.05 mg/g) and butorphanol tartrate 0.003 mg/g). Then they were perfused transcardially with 5-10 ml of warm 0.9% normal saline followed by 20 ml of 10% buffered formalin. After perfusion, the heads of the mice were stored in 10% buffered formalin overnight. The brains were removed from the skull with care to preserve the olfactory bulbs, paraﬂoccular lobes of the cerebellum and optic nerves. In total, 10 brains were prepared for MRM (magnetic resonance microscopy) imaging.
MRI Imaging (In vitro)
T2*-weighted MR images were acquired on a 17.6-T magnet @ University of Florida. 3D gradient-echo pulse sequence was used (Number of excitations = 2; TE = 7.5ms; TR = 150ms; Spectral bandwidth=42kHz; Flip angle = 25 degree; Matrix = 256x256 ; FOV = 2.4x1.2cm (47mm isotropic resolution); Imaging time = 5.5 hours). Image reconstruction was performed using Bruker Paravision software. A 3D fast Fourier transform was applied without any apodization, filtering or zero-filling.
Animal Preparation (In vivo)
Inbred C57BL/6J male mice, 12-14-week old, weighing 25-30 g were used (Jackson Laboratory, Bar Harbor, ME). All protocols for live animal experiments were approved by the Institutional Animal Care and Use Committee. For MR microscopy (MRM) the mice were initially anesthetized with an intraperitoneal injection of a mixture of Nembutal (50 mg/kg), glycopyrrolate (0.01-0.02 mg/kg) and 0.9% saline. A gas mixture of oxygen and isoflurane (1-2%) was used for anesthesia maintenance. The electrocardiogram (ECG), respiratory rate and body temperature were monitored and recorded constantly with an MR-compatible small animal monitoring system (PC_SAM, model #1025, SA Instruments). The body temperature of the mouse was kept at 36.5oC and the respiratory rate was maintained around 50-65 breath per minute (bpm) by adjusting the concentration of the isoflurane gas mixture.
MRI Imaging (In vivo)
All the in vivo mouse brain MRM images were acquired on a superconducting 9.4T/210 horizontal bore magnet (Magnex) controlled by an ADVANCE console (Bruker) and equipped with an actively shielded 11.6cm gradient set capable of providing 20 G/cm (Bruker, Billerica, MA). A birdcage radio-frequency (RF) coil (inner diameter 72 mm) was used as the transmission RF coil and a 30-mm-diameter surface RF coil as the receiver coil. We designed a new animal positioning system which provided more robust support to the animal cradle as well as better support of the RF birdcage coil against acoustic vibrations generated by the gradients and other sources. The new design also allowed us to place consistently the region of interest (i.e. the brain of the mouse) within the homogeneous center of the magnetic field. Please see our paper for more details.
T2-weighted MR data were generated with a 3D large flip angle spin echo sequence that shortens TR and the total scan time. The longest dimension of the mouse brain was set as the frequency encoding direction, with the remaining orthogonal directions sampled by phase encoding techniques. Image reconstruction was performed using Bruker Paravision software (version 4.0). A 3D fast Fourier transform was applied without any apodization, filtering or zero-filling. In all cases, 3D datasets were reconstructed into 16-bit grayscale images.
Individual Atlas Creation
Minimum Deformation (Average Shape) Atlas Creation
We constructed a minimal deformation C57BL/6J mouse brain atlas that assumes the average attributes (structure shape and MRI intensity) of the whole group. Our approach is similar to that of Kochunov and colleagues, although we modified the process by using a recursive process for more stable convergence and less bias due to different initial reference brain. Further, we used the segmented individual atlases, which contain clear defined structural boundaries that can facilitate higher registration accuracy and much less computation time. The minimal deformation atlas provides less bias as group representation.
Probabilistic Atlas Creation
A 3D volumetric probabilistic atlas was created for the 20 segmented structures of the 10 C57BL/6J brains as follows: First, the 10 individualized brain atlases were normalized to a reference space through rigid-body transformation. Then the probability, of certain structure, occupying a particular voxel in the reference brain space was computed. The atlas then was color-coded, and the variant structure occupancy at different locations can be visualized. The atlas is potentially useful in automatic segmentation procedures by providing a priori group anatomical information.
Additional Information: Group Deformation Map
The group deformation map was obtained by nonlinearly registering all the target brain atlases to the reference brain atlas and then mapping the voxel displacement onto the same 3D reference space. Because the target brains were initially registered with the reference brain through affine transformation, the deformation map represents higher order geometrical variations other than global differences such as brain location or brain sizes.
For more details, please see our paper.