Final Report Summary - HIRESBRAIN7T (High Resolution Segmentation of Brain MR Images at 7 Teslas)
A primary goal of neuroscience is to determine which specific anatomic area of the brain is associated with brain activity. High resolution in vivo MRI at 7 Tesla (7T) has opened the way to in-vivo microscopy with images of brain structures at the sub-millimetric level, showing fine details barely visible at 3T, such as the stria of Gennari or sub-regions of the Hippocampus and providing quantitative information about tissue composition at that scale. Improved resolution enhances the definition of structures, and reveals many subtle details within them, bringing us closer to the level of detail obtained in microscopy of cadaver brain.
The benefits of 7T imaging are tantalizing, but also bring in many new challenges. Even for classical scans, the new heterogeneity in tissues and the newly visible structures, as well as stronger field inhomogeneity artifact and the increased data size, present serious issues with regard to analysis. The new features observable at 7T can be subtle and require precise anatomical measurements. Automated image processing methods are obviously required in this endeavor, and it would be naive to hope that simply using current existing neuroimage analysis methods developed for 1.5T and 3T imaging will suffice. Our goal for this project was to create these new processing methods and with them explore the possibilities at the frontier of high-resolution 7T imaging.
We created in the first half of this project segmentation algorithms able to process MR images at 0.4 mm, 15 times the amount of data routinely handled by existing brain image processing software. We also introduced a model of depth within the cortex that respects the biomechanics of cortical folding.
Based on these high resolution cortical segmentations and this depth model, we have now studied in more detail the patterns of cortical architecture in various areas. We investigated different approaches to relate the T1 contrast in MR images with the organization of neurons and their axons inside the cortex.
A key question in order to draw general conclusions for multiple subjects is whether we can align their cortical anatomy. We challenged the state-of-the-art, and showed that accurate co-alignment was possible in most areas of the cortex if we used intra-cortical contrast to guide the alignment and a careful multi-scale approach.
Thanks to the higher resolution achievable at 7T and quantitative contrasts of T1 and susceptibility, we have created a comprehensive atlas of the basal ganglia, investigated the internal anatomy of its smaller nuclei and segmented even smaller structures such as the hypothalamus or the dentate nucleus.
The study of the brain's activity is intimately linked to its vascularisation. We have developed new methods to extract the venous vasculature from MR images of quantitative susceptibility and model the local oxygenation within each vein before it enters the cortex.
Finally, we have answered the recent calls to make science more open and reproducible by making our best data sets freely available to the public along with all our software tools:
http://openscience.cbs.mpg.de/bazin/7T_Quantitative/
http://www.nitrc.org/projects/atag/
http://www.nitrc.org/projects/atag_mri_scans/
http://www.nitrc.org/projects/cbs-tools/
http://www.cbs.mpg.de/institute/software/cbs-hrt/index.html
The benefits of 7T imaging are tantalizing, but also bring in many new challenges. Even for classical scans, the new heterogeneity in tissues and the newly visible structures, as well as stronger field inhomogeneity artifact and the increased data size, present serious issues with regard to analysis. The new features observable at 7T can be subtle and require precise anatomical measurements. Automated image processing methods are obviously required in this endeavor, and it would be naive to hope that simply using current existing neuroimage analysis methods developed for 1.5T and 3T imaging will suffice. Our goal for this project was to create these new processing methods and with them explore the possibilities at the frontier of high-resolution 7T imaging.
We created in the first half of this project segmentation algorithms able to process MR images at 0.4 mm, 15 times the amount of data routinely handled by existing brain image processing software. We also introduced a model of depth within the cortex that respects the biomechanics of cortical folding.
Based on these high resolution cortical segmentations and this depth model, we have now studied in more detail the patterns of cortical architecture in various areas. We investigated different approaches to relate the T1 contrast in MR images with the organization of neurons and their axons inside the cortex.
A key question in order to draw general conclusions for multiple subjects is whether we can align their cortical anatomy. We challenged the state-of-the-art, and showed that accurate co-alignment was possible in most areas of the cortex if we used intra-cortical contrast to guide the alignment and a careful multi-scale approach.
Thanks to the higher resolution achievable at 7T and quantitative contrasts of T1 and susceptibility, we have created a comprehensive atlas of the basal ganglia, investigated the internal anatomy of its smaller nuclei and segmented even smaller structures such as the hypothalamus or the dentate nucleus.
The study of the brain's activity is intimately linked to its vascularisation. We have developed new methods to extract the venous vasculature from MR images of quantitative susceptibility and model the local oxygenation within each vein before it enters the cortex.
Finally, we have answered the recent calls to make science more open and reproducible by making our best data sets freely available to the public along with all our software tools:
http://openscience.cbs.mpg.de/bazin/7T_Quantitative/
http://www.nitrc.org/projects/atag/
http://www.nitrc.org/projects/atag_mri_scans/
http://www.nitrc.org/projects/cbs-tools/
http://www.cbs.mpg.de/institute/software/cbs-hrt/index.html