Automated MRI could improve stroke research by making brain tissue measurements more consistent across large animal studies, researchers say.
The system was tested on scans from more than 2,000 mice and rats across six academic research centres, with measurements closely matching those made by human experts.
By reducing differences between thousands of preclinical scans from different sources, the automated system could improve the early-stage evaluation of new treatments.
The system was developed by researchers at the Mark and Mary Stevens Neuroimaging and Informatics Institute at the Keck School of Medicine of USC.
It was created for the US National Institutes of Health-sponsored Stroke Preclinical Assessment Network, or SPAN, which tests potential treatments for acute ischaemic stroke in coordinated animal studies before the most promising therapies advance towards clinical trials.
The Stevens INI serves as the informatics and data core for SPAN.
Co-first author Kirsten M. Lynch, assistant professor of research neurology, said: “Before a potential treatment can be tested in people, researchers need confidence that its effects have been measured rigorously and consistently.
“This pipeline gives us an objective and scalable way to assess brain injury across a large research network.”
An ischaemic stroke occurs when a blocked blood vessel cuts off blood and oxygen to part of the brain.
Although many treatments have shown promise in laboratory studies, few have proved effective in patients.
Differences in how early-stage studies are carried out and analysed can make results difficult to reproduce or compare across research centres.
Scientists have traditionally measured damage after an experimental stroke by removing the brain, staining tissue sections and manually outlining injured areas.
The process can distort tissue, depends partly on individual judgement and captures only one point in time.
MRI allows researchers to scan the same animal more than once, tracking early injury and swelling as well as tissue loss that develops later.
However, analysing thousands of scans collected using different equipment creates another challenge.
“The scale of SPAN made automation essential,” said Ryan Cabeen, a computational scientist at the Stevens INI who led development of the imaging biomarker platform and was co-first author of the study.
“We needed a method that could process thousands of scans while applying the same rules to every image, regardless of where the data were collected.”
The automated pipeline checks image quality, adjusts for differences between scanners, identifies the brain and measures injured tissue, swelling, displacement and longer-term tissue loss.
It uses a deep-learning model to separate the brain from surrounding bone, muscle and other tissue.
The system then applies transparent, rule-based methods to identify stroke damage rather than relying entirely on artificial intelligence.
“We wanted researchers to understand how the results were produced,” Lynch said.
“A method can be highly automated without becoming a black box. The combination of deep learning and transparent image-processing rules gave us both robustness and interpretability.”
The study involved 2,442 mice and rats.
More than 2,200 were scanned two days after an experimentally induced stroke, while 1,750 received another scan about one month later.
When the automated results were compared with injuries manually outlined by imaging experts, the measurements showed extremely close agreement.
The system produced results about as consistently with a human reviewer as two human reviewers did with each other.
It also reduced variation linked to different scanners and imaging environments, allowing the same analysis approach to work across all six research centres.
The software successfully processed the vast majority of scans despite differences in species, equipment, magnetic field strength and image quality.
Researchers have made the software and MRI data publicly available so other groups can reproduce the findings and adapt the pipeline for future studies.
The tool was designed for standardised animal models and is not intended to analyse the more varied stroke injuries seen in patients.
However, its framework could be expanded to include additional imaging methods and measures of brain tissue outcomes.
“Large, collaborative studies require tools that produce reliable results across institutions,” said Arthur W. Toga, director of the Stevens INI and a co-author of the study.
“By combining advanced imaging, artificial intelligence, data harmonisation, and high-performance computing, this work provides a reproducible foundation for evaluating which experimental stroke treatments have the greatest potential to move toward clinical testing.”
