Model-Based Parametric Mapping, with Code

In a paper published recently in NMR in Biomedicine, CAI2R’s Florian Knoll describes a new way to reconstruct parametric maps from scan data in diffusion MRI. The method enables shorter scan times while retaining satisfactory information, and can be used by “anybody who is interested in accelerated imaging and parametric mapping,” said Dr. Knoll in an interview. Indeed, anyone can download the associated MATLAB code from our resource page.
Results from the Model-Based Diffusion Tensor Imaging (Model DTI) Reconstruction
compared to other methods.  
The technique does away with conventional two-step process in which radiologists first reconstruct images, and then calculate quantitative information based on these. Dr. Knoll wondered, “if I’m ultimately interested in diffusion, why do I separate the two steps, why not do everything at once?”
The technique can be used by anybody interested in imaging and parametric mapping.
He has discovered that both steps can be merged. The solution, which incorporates concepts from the fields of inverse problems and mathematical optimization, relies on deliberately constraining the processing of raw scan data (in this case, diffusion data). In other words, scientists know a priori that the measured domain must have particular properties and “impose that knowledge during reconstruction,” explained Dr. Knoll, referring to the method as model-based, where model means set of substantiated assumptions. Combining the stages is key: “This way you make use of all available information,” which would otherwise be lost between steps, he added.
Working with Jose Raya, an NYU investigator focusing on diffusion imaging in cartilage, Dr. Knoll tested the technique both in phantoms and in vivo, obtaining results rivaling and at times surpassing current gold standard methods (Dr. Raya is using the approach to develop early detection of osteoarthritis). But nothing’s free—shorter scans come at the cost of longer reconstructions due to the numerically challenging nature of the solution.
The technique applies universally: “In our case the parameter is diffusion information, but it could be anything, it could be any kind of biophysical parameter,” stressed Dr. Knoll.


Latest Updates

04/05/2021 - 09:00
03/24/2021 - 17:06

Philanthropic Support

We gratefully acknowledge generous support for radiology research at NYU Langone Health from:
• The Big George Foundation
• Bernard and Irene Schwartz

Go to top