Loading evaluation data…
Studies demonstrate reliable detection of core Alzheimer disease markers, including hippocampal atrophy and white matter hyperintensities, on portable low-field systems, with promising correspondence to conventional high-field MRI findings.
small dementia cohorts and predominantly MS-based WMH validation limit the generality of reliable marker detection
Susceptibility-based markers, such as microbleeds, which are crucial for vascular characterization and ARIA-H monitoring, remain poorly detectable on low-field MRI.
limited clinical validation includes failed ARIA-H detection in a larger sample and restricted susceptibility-sensitive imaging capability
An additional unresolved issue is the sensitivity of low-field MRI to longitudinal change, as strong cross-sectional correlations with high-field MRI might not translate into equivalent performance for detecting intra-individual change.
cross-sectional validation does not establish sensitivity to within-person change
Despite low resolution and signal-to-noise ratios, innovations such as segmentation algorithms and super-resolution methods extend the capability of low-field MRI, although validation across diverse populations is still required.
processing studies demonstrate added capability but lack validation across sufficiently heterogeneous populations
Derived from the full evaluation — not a separate score.
Strengths
The review brings regional atrophy, WMHs, diffusion imaging and ARIA into a single clinical and research framework. Figure 6 connects these capabilities to access, population representation and workforce requirements.
↳ High-field versus low-field MRI in neurology; Fig. 6
The marker summary distinguishes structural correspondence from unestablished longitudinal sensitivity and poor microbleed detection. The clinical-trial discussion consequently proposes remote ARIA-E triage or surveillance rather than comprehensive replacement of standard MRI.
↳ Summary of markers on low-field MRI; Potential in clinical trials
The discussion considers maintenance failures, training, connectivity, AI governance and cost per scan. It explicitly recognizes that lower capital costs do not establish cost-effectiveness or sustainable use.
↳ Economic perspectives; Barriers to adoption; Methodological challenges
Limitations
No search or source-selection process is described for the empirical synthesis. Readers therefore cannot determine how the included evidence was assembled.
↳ Introduction and article-wide narrative synthesis
The Key points describe reliable detection of core markers, while the two dementia-specific samples contain 54 and 55 participants and one report is a preprint. Much of the cited WMH validation comes from MS rather than dementia cohorts.
↳ Key points, bullet 2; Grey matter volume; White matter hyperintensities
Longitudinal sensitivity and performance across diverse populations remain unestablished, while mobile studies do not establish performance under poor-road or unreliable-power conditions. No cost-effectiveness analysis supports routine dementia deployment.
↳ Box 1; Summary of markers on low-field MRI; Economic perspectives; Barriers to adoption
The review's contribution is its dementia-focused integration of marker performance, clinical pathways and global implementation requirements. Quantitative validation findings and negative ARIA-H evidence support a complementary role for low-field MRI, while the absence of a study-selection account limits the synthesis's methodological transparency. The small dementia cohorts and largely MS-based WMH validation warrant more restrained headline confidence than 'reliable detection' suggests. Box 1 and the clinical-trial discussion nevertheless preserve appropriate boundaries by acknowledging unknown longitudinal sensitivity and proposing indication-specific uses.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.7
Confidence highThe score reflects a dementia-specific synthesis linking imaging markers to clinical, research and equity considerations beyond general low-field MRI reviews. Its contribution is an organizing framework rather than new empirical validation.
The marker-by-marker argument integrates quantitative findings and countervailing evidence, including failed ARIA-H detection and untested longitudinal sensitivity. The absence of an account of study identification or selection is already priced as a 0.3667-point reduction from mean merit.
Sectional progression and Box 1 make the distinction between current capabilities and proposed applications followable. The score also reflects stronger headline wording than the small dementia-specific validation cohorts warrant.
Technical limitations are connected to clinical scope, AI governance, economics and implementation requirements rather than simply listed. Comparative positioning against CT, blood biomarkers and mid-field alternatives is less developed.
Lower confidence on Reporting, Positioning — domain match limited.
Caveats4 of 4 checks
The reported numbers and statistics are consistent with the stated results.
The dementia-growth projection periods differ between the Introduction and Fig. 1 without reconciliation. Competing interests are disclosed.
Conflict of interest: declared
183 references in manuscript 174 of 180 checkable references found in an index 3 are books, websites or datasets — counted, but not index-checkable
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
Structural imaging has proof-of-concept support, but dementia-specific deployment remains constrained by small validation samples and unestablished longitudinal performance. Missing cost-effectiveness evidence further limits readiness for routine care.
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
decreased ACD was associated with greater AD pathological burden
Supported2026 · Clinical medicine
plasma p-tau217 demonstrated a positive percent agreement of 84.5 % (95 % CI: 72.6–92.7 %) and a negative percent agreement of 87.5 % (95 % CI: 73.2–95.8 %).
Tentative