Skip to main content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Deep Learning for MRI Reconstruction and Analysis

Summary

Healthcare is big part of the national economy.  Medical Imaging has emerged over the past several decades as a critical diagnosis and therapeutic tool.  However, the metrology in these tools is not tied to the SI system of units.

From a technical perspective, medical imaging generates very large data sets.  AI systems, based on Deep Learning (DL), have emerged as key computational approaches for the quantitative analysis of these images.

The goal of the project is to initiate the development of the metrology and standards infrastructure to ensure that medical DL-based systems are (1) trained on validated physics-based data and (2) provide reliability, accuracy, and explainability.

Description

The project is proceeding in three directions.

  1. Creating a new MRI reference artifact designed to assess geometric distortion using NIST’s MRI scanner.  The artifact will be small enough to fit within the scanner with sufficient clearance to allow for variation in positioning within the scanner.
    We plan to use this artifact to generate a data set that will then be used to assess the accuracy of existing DL-based approaches.
  2. Develop a targeted MRI simulation capability with attention focusing on physics, scan types, and analysis pathways that are relevant to the MRI reference artifact.
    We plan to use this capability to reproduce the MRI component of Zhu, et al., and to document our efforts.
  3. Prototype a DL-based MRI Inversion problem.  This will start by reproducing the results from an existing implementation, AUTOMAP.

 

Created March 12, 2019, Updated March 15, 2019