By leveraging machine learning tools, researchers have measured markers of Alzheimer’s disease on a portable MRI with accuracy matching that of standard MRI.
Globally, approximately 139 million people are expected to have Alzheimer’s disease by 2050.
Magnetic resonance imaging (MRI) is an important tool for identifying changes in brain structure that precede cognitive decline and progression with disease; however, its cost limits widespread use.
Now, a new study demonstrates that a simplified, low magnetic field (LF) MRI machine, augmented with machine learning tools, matches conventional MRI in measuring brain characteristics relevant to Alzheimer’s.
The findings highlight the potential of the LF-MRI to help evaluate those with cognitive symptoms.
“To tackle the growing, global health challenge of dementia and cognitive impairment in the aging population, we’re going to need simple, bedside tools that can help determine patients’ underlying causes of cognitive impairment and inform treatment,” said senior author W. Taylor Kimberly, chief of the Division of Neurocritical Care in the Department of Neurology at MGH.
“To do this, we brought a research team together to design a patient-centred approach to brain imaging to improve access and convenience, reduce costs, and streamline cognitive evaluations.”
The research team behind the project included clinical researchers, MRI physicists, health system delivery experts, and AI experts.
Investigators at MGH have explored LF-MRI for several years as an alternative to traditional high-field (HF) MRI, which provides detail-rich, cross-sectional images of the body by generating powerful magnetic fields.
HF-MRI machines are expensive, require designated imaging centres/suites for operation, and are frequently absent from low-resource settings in the U.S. and around the world.
In contrast, LF-MRI machines use magnetic fields 50 times weaker than those required for conventional MRI. This makes LF-MRI scanners smaller and portable, requiring just one electrical outlet for operation, but also results in reduced image quality.
To improve LF-MRI image quality and make it easier to use in the clinic, the researchers leveraged artificial intelligence (AI) tools. The researchers created artificial datasets and matched HF- and LF-MRI scans from both healthy people and patients with neurological conditions.
They used these datasets to train an algorithm to recognise features relevant to Alzheimer’s on LF-MRI, including the size of brain structures like the hippocampus (the brain’s memory centre), and white matter hyperintensity (WMH) regions, which can indicate neurodegeneration or blood vessel problems.
When they tested their method in 54 patients with mild cognitive impairment or AD-related dementia, the AI-based LF-MRI scans closely matched traditional, HF-MRI measurements of the hippocampus and white matter volume.
The new approach will require regulatory clearance and new clinical protocols, but it holds the promise of expanding neuroimaging in settings with limited MRI access.
Beyond advancing Alzheimer’s diagnosis, LF-MRI may help streamline care for Alzheimer’s patients who require MRI monitoring during treatment with novel Alzheimer’s drugs. The portable LF-MRI could also be used in emergency rooms, community health centres, or ambulatory units, especially for patients who have experienced or are at risk of stroke.
“Access to traditional MRI is not evenly distributed and not available to everyone,” Kimberly said.
“We envision a future where a person with cognitive complaints visiting a primary care physician, geriatrician or neurologist can get a brain scan, a blood test and a cognitive test, all in a single visit. Low-cost, easier to deploy technology can help provide information to clinicians, right at the bedside.”
