NIH is increasing its investment in large, multi-center brain MRI studies via projects such as the BRAIN initiative. The success of these studies depends on the quality of MRIs and the resulting image measurements.
Many neuroscience studies rely on in-house procedures that combine automatically generated scores with manually guided checks. This is often done by combining several software systems that are not designed to support Quality Assurance (QA) or Quality Control (QC) processes. Furthermore, collaborative technologies that assist a distributed team to implement rapid and more accurate Quality Control activities are missing in multi-site QC workflows. Therefore, we are developing Medical Image Quality Assurance (MIQA) software. MIQA represents a design that facilitates collaboration and sharing. It also incorporates a state-of-the-art deep learning system to help focus reviewers’ attention to potentially problematic scans.
· simplify creation of custom QC workflows in
compliance with study requirements
· provide core functionality for performing QC of
· automatically generate documentation compliant
with the FAIR principle
FAIR means making scientific results Findable, Accessible, Interoperable, and Reusable.
We designed MIQA for medical imaging QA/QC from the ground up. Users will be able to configure workflows that reflect the specific requirements of medical imaging studies. We also aim to minimize the time spent on labor-intensive operations, such as visually reviewing scans. Our deep learning techniques will include incremental learning strategy that continuously improves the machine-generated QC score.
The usefulness of this unique QC system will be tested through user studies. One of them is the multi-center National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA) project. The Institute for Clinical and Translational Science at the University of Iowa will provide further testing.
A work-in-progress version of MIQA has been deployed at Stanford. The main purpose so far is to get early feedback from our clinical collaborators. Current developments are user interface reorganization. The UI now includes display of AI-produced scan rating. The next step is addition of multi-user roles and permissions.
This project is funded by NIH grant R44 MH119022 from the National Institute of Mental Health.