In the realm of dementia care, a remarkable transformation is underway. Explore how portable, AI-enhanced magnetic resonance imaging (MRI) systems are shattering barriers in the diagnosis of Alzheimer’s disease, paving the way for early detection and enhanced global reach.
A recent study published in Nature Communications has optimized the use of portable low-field MRI (LF-MRI) acquisition and devised a machine learning pipeline. This innovative approach aims to estimate brain morphometry and quantify white matter hyperintensities (WMH), crucial elements in the diagnosis of Alzheimer’s disease.
Alzheimer’s is a progressively debilitating neurodegenerative ailment that impairs memory, cognition, and behavior. Pathologically, it is marked by the accumulation of amyloid-β (Aβ) and the formation of neurofibrillary tangles within the brain. Over time, the build-up of these proteins triggers adverse changes in brain structure and escalates vascular injury, observable through quantifiable brain atrophy and WMH respectively.
Typically, the pre-symptomatic stage of AD can span 10 to 20 years. This protracted period likely accounts for why a staggering 75% of individuals with dementia remain undiagnosed for extended durations. With the advent of anti-amyloid therapies, the need for early detection of AD and mild cognitive impairment (MCI) has become even more pressing, as timely diagnosis maximizes treatment efficacy.
AD diagnosis currently hinges on cognitive testing, which evaluates the Aβ and phosphorylated tau burden via fluid biomarkers, positron emission tomography (PET), and MRI. Clinicians glean insights into brain structure and integrity changes from multi-contrast MRI. Key imaging indicators such as generalized and hippocampal atrophy assist physicians in tracking disease progression and cognitive decline.
While neuroimaging is undeniably invaluable in AD and MCI diagnosis and management, its limited availability, both locally and globally, has contributed to underdiagnosis. Previous research spotlighted the development of portable LF-MRI, touting its potential to boost accessibility and enhance diagnosis of diverse neurodegenerative conditions. However, its reduced magnetic field strength has been a stumbling block, leading to a lower signal-to-nonoise ratio (SNR) and compromised image resolution.
The present study tackled the LF-MRI limitations for AD and MCI diagnosis head-on by developing machine learning tools capable of automatically quantifying brain morphometry and white matter lesions.
An imaging pipeline was constructed to precisely quantify brain volumes. The refined super-resolution and contrast synthesis technique, known as LF-SynthSR, was optimized to augment LF image resolution for subsequent segmentation (SynthSeg). For instance, hippocampal volumes derived from LF-MRI closely mirrored those from high-field MRI, with an Absolute Symmetrized Percent Difference (ASPD) of 2.8% and a Dice similarity coefficient of 0.87.
This methodology established the optimal LF acquisition parameters for accurate quantification. It also enabled the measurement of WMH burden (WMH-SynthSeg) through automated segmentation of WMH lesions from T2 fluid-attenuated inversion recovery (FLAIR) images acquired at LF. The study validated LF-SynthSR, SynthSeg, and WMH-SynthSeg using a prospective cohort of patients diagnosed with MCI or AD.
To build this imaging pipeline, participants from three cohorts were enlisted. They underwent MRI scans on a portable, low-field 0.064 T MRI, as well as a high-field, conventional scan at 1.5–3 T. The first cohort comprised 20 healthy individuals (10 males and 10 females) with no history of neurological disease or memory complaints. The second cohort consisted of 23 participants (11 males and 12 females) possessing at least one vascular risk factor but no current neurologic complaints or prior memory disorder history. The third cohort included 54 individuals (32 males and 22 females) diagnosed with MCI or AD. These participants adhered to an LF-MRI imaging protocol incorporating T1w, T2w, and FLAIR sequences.
Although LF-MRI images lacked the resolution required for automatic segmentation with high-field software analysis tools, they were super-resolved (SR) into 1 mm isotropic T1-weighted (T1w) magnetization-prepared rapid gradient-echo (MP-RAGE)-like images. The study determined that isotropic voxel sizes of ≤3 mm enhanced segmentation accuracy, yielding ASPD values under 5% for hippocampal volumes. Moreover, the refinement of the LF-SynthSR v2 pipeline improved automated segmentation accuracy, amplifying the usability of low-field imaging applications.
In the first cohort, the accuracy of automated segmentation was gauged by comparing AD-relevant segmentation volumes of the hippocampus, lateral ventricle, and whole brain generated from the original LF-SynthSR and LF-SynthSR v2 against conventional high-field (HF) MRI acquired at 3 T.
Comparing LF-SynthSR v2 with LF-SynthSR v1 led to an improvement in lateral ventricle volume accuracy. Image acquisition time ranged from 1:53 to 9:48 minutes, contingent on voxel size and sequence. The study also found that isotropic voxel sizes of ≤3 mm bolstered segmentation accuracy, especially in the low SNR environment of LF-MRI. The accuracy of brain morphometry was found to be influenced by voxel size and geometry. Furthermore, the LF-SynthSR v2 segmentation pipeline was validated against HF T1w MP-RAGE segmentations derived from the FreeSurfer segmentation tool ASEG.
WMH lesions, often a consequence of axonal loss or cerebral small vessel disease, were prevalent among patients with cognitive impairment. These were quantified using WMH-SynthSeg. The utilization of these findings on FLAIR as T2 hyperintense lesions and the automated quantification of such lesions elevated the AD diagnosis and monitoring capacity of LF-MRI.
This study harnessed machine learning to generate WMH lesion volumes (WMHv) from LF-FLAIR images using WMH-SynthSeg. This strategy enabled simultaneous segmentation of WMH T2 FLAIR lesions alongside prior brain morphometry. The WMH volumes correlated robustly with manual annotations and high-field imaging standards.
Based on WMHv generated by WMH-SynthSeg, the machine learning tool was validated as it could accurately detect patients with MCI, AD, and those who were cognitively normal (CN).
The current study conclusively demonstrated that LF-MRI, when paired with machine learning tools, has the prowess to diagnose patients with AD or MCI. Looking ahead, this device holds promise for assessing its ability to detect neurodegenerative tauopathies and vascular dementia. Its portability, affordability, and automated analysis pipeline signify immense potential for bridging global diagnostic gaps.
Related topics
- Study Reveals Alarming Rates of Depression and Anxiety Among Australian Adolescents
- Chemogenetic Activation Finds Sex-Specific Stress Resistance Circuits
- Open-World Games Emerge as a Stress Buster for Postgraduates
