Projects currently offered to current Harvard-MIT Health Sciences and Technology (HST) students

Generative models of pathology for clinical brain MRI analysis.

Our group has recently developed machine learning techniques for image analysis of brain MRI scans of any resolution and contrast, which has finally enabled the analysis of scans from our hospital PACS. While these techniques work remarkably well, we have so far trained them with relatively normal anatomy. The goal of this project is to learn to synthesize pathology during training (e.g., tumors, strokes) in order to enable the application of our methods in the wild.

Super-resolution of diffusion MRI with domain randomization and implicit neural representations (collaboration with Dr. Anastasia Yendiki)

Super-resolution of diffusion MRI enables the tracking of the thinner and more convoluted fiber tracts in the human brain. The goal of this project is to generally learn to super-resolve diffusion MRI data jointly in the spatial and angular domains, using a combination of domain randomization and implicit neural representations. Combining such algorithms with the unique ex vivo data generated with the newly developed Connectome 2.0 scanner at the Martinos should greatly improve our ability to analyze in vivo scans of living people.


Postdoc job opportunity: Postdoc in multimodal and multiscale registration for building a transcriptomic atlas of the human brain using machine learning (NIH BRAIN Initiative)

A postdoctoral position in Multimodal and multiscale registration for building a transcriptomic atlas of the human brain using machine learning (funded by the NIH BRAIN Initiative) is available for application at the Laboratory for Computational Neuroimaging, Martinos Center for Biomedical Imaging, Massachusetts General Hospital (MGH) and Harvard Medical School.

The Laboratory for Computational Neuroimaging (LCN) is seeking to fill a postdoctoral position in image analysis as part of an exciting 5-year NIH BRAIN Initiative consortium project titled: “A transcriptomic atlas of the human brain”. The project is a collaboration between 18 different institutions in three different continents: MGH, MIT, the Allen Institute, Princeton, University of Washington, UPenn, Arizona State, Baylor, and Washington University (USA); Karolinska Institute (Sweden); McGill (Canada); Inserm (France); RIKEN (Japan); Tubingen (Germany); Szeged (Hungary); Amsterdam and Leiden (The Netherlands); and the EMBL's European Bioinformatics Institute (UK). The goal of the project is to create a comprehensive cell atlas of the human and closely related non-human primate brain, using high-throughput single cell molecular, spatial and cell phenotyping assays to map the structural and functional organization of the primate brain.  The result will be a transformative resource for linking brain structure, function, cellular makeup and genomic characteristics that provide a powerful new foundation to accelerate understanding of human brain function and disease.

We are seeking a talented and enthusiastic postdoctoral researcher to develop machine learning and image analysis methods for this project. The postholder will work with Dr. B. Fischl. Dr. J.E. Iglesias, other LCN members, and collaborators in the consortium to develop methods for: (i) 3D photography and registration with 3D surface scanning data; (ii) mapping in vivo atlases to histology to guide sampling for transcriptomic analysis at the Allen Institute; (iii) optimization and image analysis of ultra-low-field MRI scanning at UW; and (iv) building a transcriptomic atlas for use in in vivo studies.

Applicants should have:

- A Ph.D. degree in computer vision, neuroimaging, or biomedical engineering/image analysis. Applicants with degrees in related areas (e.g., Engineering, Physics, Applied Mathematics) will also be considered. Applicants close to submission of their PhD thesis will be considered as well.
- Strong mathematical and problem-solving abilities.
- Expertise in programming, at least with high-level languages / packages (Python, PyTorch, Tensorflow).
- An established publication track record.
- Ideally, experience with one/some of the following: neuroimaging, deep learning, image registration, or geometric analysis of meshes and point clouds.

The position is available as of October 2022, but the start date is flexible. Applications will be reviewed as they arrive. The duration of the contract is two years, with possibility of further extension.

For applications, please send the following documents to
- A cover letter (maximum 1 page)
- CV (maximum 4 pages)
- List of publications