Multimodal, high-resolution modeling of the thalamus for neuroimaging studies: application to dyslexia

Project Information

THALAMODEL

Grant agreement ID: 654911

DOI

10.3030/654911

Project terminated on 31 August 2016

EC signature date 2 March 2015

Start date 1 June 2015

End date 31 May 2017

Funded under

• EXCELLENT SCIENCE - Marie Skłodowska-Curie Actions

Total cost

€ 170 121,60

EU contribution

€ 170 121,60

Coordinated by

BCBL BASQUE CENTER ON COGNITION BRAIN AND LANGUAGE
Spain

Periodic Reporting for period 1 - THALAMODEL (Multimodal, high-resolution modeling of the thalamus for neuroimaging studies: application to dyslexia)

Reporting period: 2015-06-01 to 2017-05-31

Summary of the context and overall objectives of the project

The thalamus is a cerebral structure that plays the role of main relay station of the human brain. It is involved in a number of important functions such as language and regulation of consciousness and sleep. These functions are distributed across dozens of thalamic nuclei, which are connected to the rest of the brain very differently. However, most neuroimaging software tools still consider the thalamus as a whole.

The goal of this project was to build a map of the human thalamus at the nucleus level, using high quality, ex vivo imaging data (i.e. from dead brains). Since subject motion is not a problem in ex vivo imaging, we scanned the samples for long periods of time in a MRI scanner to obtain images with high resolution. Also, we performed histological analysis of the tissue, i.e. sectioning, staining and high-resolution digitization in order to see boundaries between thalamic nuclei that would otherwise be invisible – even in high resolution MRI.

The thalamic nuclei where manually delineated by an expert neuroanatomist on the high resolution ex vivo images. We then used these delineations to build an atlas (i.e. a probabilistic map of anatomy) of the thalamus. Using software tools developed during this project, the built atlas can be used to analyze in vivo brain MRI scans (i.e. from living subjects) and study the thalamic nuclei in living people. The atlas and companion tool will be made publicly available through the widespread neuroimaging package FreeSurfer (over 20,000 users worldwide), and will allow scientists around the world to disentangle the contribution of the different thalamic nuclei to the effects observed in their studies. For example, the tool will enable us to test whether the effects of dyslexia on the left thalamus reported in previous studies are confined to two nuclei known as the LGN and MGN or shared by other nuclei as well. Analyses like these are not possible with current tools, which consider the thalamus as a whole.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

"First, we scanned the brain samples at the BCBL. Since the MRI scanner is not designed for ex vivo samples, it created an artifact in the images known as the Venetian blind. We designed a method to correct this artifact, published it in a conference article, and presented it in an international MRI conference. The software implementation can be downloaded from the project website.

The histological analysis (slicing, staining, digitization) of the samples was carried out by our collaborator Dr. Ricardo Insausti at the University of Castilla – La Mancha (UCLM). Dr. Insausti has also been in charge of the manual delineation of the thalamic nuclei on the digitized histological images, which is still ongoing (three processed cases out of six), and is expected to be finished in the next few months.

We have also developed a method to recover the 3D structure that was lost when slicing the brain tissue (“histology reconstruction”). Our method relies on photographs of the tissue that were taken during slices (“block-face photographs”) and also the MRI data. We have successfully estimated the deformations with this method, and we will use them to recover the 3D structure of the delineations as soon as Dr. Insausti has finished them. We have submitted an abstract describing this work to an international conference (Organization for Human Brain Mapping, OHBM). We will extend this abstract to a full journal article when all the data are available.

Using the manual delineations that we have for now, we have used our previously developed algorithm to construct a preliminary atlas for the thalamic nuclei (see attached screenshot). We have also developed a companion tool that applies the atlas to the automated labeling (“segmentation”) of the thalamic nuclei, also based on methods that we have developed in the past (see attached image). The atlas and companion segmentation are part of the OHBM abstract and of the future journal article, too.

We have also developed an extension of our segmentation algorithm to longitudinal data, i.e. multiple scans from the same subjects at different time points. Our proposed method was published in the journal Neuroimage, and its version for the hippocampus has been made publicly available as part of FreeSurfer (we will adapt it to the thalamus as soon as the thalamic atlas has been built).

It is also worth mentioning the output of lines of research parallel to this Marie Curie project in which I have been involved as a collaborator (see publications below). This includes work with partners at the BCBL, the Danish Technical University (DTU), Harvard, MIT, and Singapore.

Finally, an ERC Starting Grant submission extending (and written during) this MSCA project has been funded by the European Research Council. The project (grant #677697, BUNGEE-TOOLS) aims to build a set of next-generation computational tools that will enable neuroimaging studies to take full advantage of the increased resolution of modern MR technology, opening new opportunities of research in neuroscience.
"

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

The impact of the project from the scientific perspective can be summarized as follows:

• The longitudinal tool that we have developed is currently distributed as part of FreeSurfer, hence enabling its over 20,000 users around the world to carry out more robust and sensitive longitudinal analyses.
• Upon completion of the manual segmentations, the thalamic atlas will be built and distributed along with the companion segmentation tool as part of FreeSurfer, as well. We expect the impact of this tool to be very large, since it will enable the neuroimaging community to study the thalamus at the substructure level, disentangling the contributions from its different nuclei.
• When the atlas is ready, I will collaborate with Dr. Paz-Alonso to process his dyslexia dataset, and explore the role of the thalamus in this disorder in monolingual and bilingual populations.

In terms of communication and results dissemination, these happened across different channels:

• Talk for the general public at the Science Museum in San Sebastian, part of the Brain Week 2016 (“Analyzing the brain in HD”).
• The project website (http://www.jeiglesias.com/thalamodel) part of the author website, which includes news, articles and code releases.
• Tweets and Facebook posts by the BCBL.
• Two presentations at internal BCBL meetings.
• Invited lectures: Institute of Biomedicine of Seville, Imperial College, University College London.
• Conference presentations: International Society for Magnetic Resonance in Medicine (ISMRM), 2015, and IEEE International Symposium on Biomedical Imaging (ISBI), 2016.
• Open access journal articles: “Bayesian longitudinal segmentation of hippocampal substructures in brain MRI using subject-specific atlases” (1st author) and “Fast and Sequence-Adaptive Whole-Brain Segmentation Using Parametric Bayesian Modeling” (2nd author), both published in NeuroImage.
• The longitudinal segmentation method has been released in open source format as part of FreeSurfer. The simultaneous bias field and slab boundary artifact methods have been made publicly available at the project website, as well.

Coronal plane of the probabilistic atlas

Similar coronal plane from a segmented in vivo MRI scan