Ultra‐fast MRI for dementia diagnosis and treatment eligibility: A prospective study

Miguel Rosa‐Grilo, Haroon Chughtai, David L. Thomas, Daniel C. Alexander, Millie Beament, Christopher R S Belder, Hans Rolf Jäger, Emma Lim, Nicholas Magill, Dermot Mallon, Jennifer M. Nicholas, Owen Nicholas, F Urbán, Geoff J.m. Parker, Frederik Barkhof, Catherine J. Mummery, Nick C. FoxAlzheimer 2025 · DOI

TL;DR

This study introduces and validates an "ultra-fast" MRI protocol for dementia diagnosis that significantly reduces scan time (by 63%) while maintaining diagnostic reliability comparable to standard protocols. This advancement could dramatically improve access to crucial MRI scans for dementia patients, accelerating diagnosis and enabling timely eligibility assessment for new disease-modifying therapies.

What Problem Does This Paper Address?

Diagnosing dementia, particularly Alzheimer's disease (AD), relies heavily on Magnetic Resonance Imaging (MRI) for assessing brain atrophy and identifying markers like microhemorrhages. Furthermore, eligibility for emerging amyloid-lowering therapies in AD mandates MRI. However, the lengthy acquisition times of standard MRI protocols create a significant bottleneck, limiting scanner availability, increasing patient burden, and delaying diagnoses and treatment initiation. This paper tackles the challenge of developing and validating a substantially faster MRI protocol that can overcome these access barriers without compromising diagnostic accuracy or reliability.

Methods

The core innovation lies in the use of wave-controlled aliasing in parallel imaging (wave-CAIPI), an advanced parallel imaging technique that allows for significantly higher acceleration factors. The researchers developed a fast MRI protocol incorporating four key sequences (T1-weighted, T2-weighted, FLAIR, and SWI) using wave-CAIPI. This fast protocol was compared to the standard-of-care clinical protocol in a prospective, real-world study.

The study design involved:

  • Image Acquisition: Participants underwent both the clinical and the fast MRI protocols in the same session on a Siemens MAGNETOM Prisma Fit 3T system.
  • Image Evaluation: Three experienced neuroradiologists independently assessed de-identified image datasets from both protocols. They were blinded to the scan type and patient information, focusing on diagnostic assessments, visual scale ratings (MTA, Koedam, Fazekas), microhemorrhage scoring, and eligibility for amyloid-lowering DMTs.
  • Reliability Analysis: The primary method for evaluating reliability was the Gwet's AC₁ statistic (and its weighted version AC₂), chosen for its robustness in situations with low category prevalence. The study focused on:
  • Inter-rater reliability: Agreement among different neuroradiologists on the same scan type (clinical vs. fast).
  • Intra-rater reliability: Agreement of a single neuroradiologist assessing different scan types (clinical vs. fast) for the same patient.
  • Comparison Benchmark: The reliability of the fast protocol was compared against the reliability of the standard clinical protocol, and also against the reliability of different raters assessing the same clinical scan.

Datasets

The study utilized data from 92 participants recruited from a specialist cognitive disorders service at the National Hospital for Neurology and Neurosurgery, UK. These individuals were undergoing MRI as part of their routine diagnostic workup for cognitive concerns. The dataset comprised both clinical and fast MRI scans from these participants. No new benchmarks were introduced; the study aimed to validate a new protocol against existing standards.

Key Insights and Ideas

  • Wave-CAIPI is a Game-Changer for Dementia MRI: The study demonstrates that wave-CAIPI can be effectively integrated into a multi-sequence dementia protocol to achieve substantial time savings without sacrificing diagnostic quality.
  • Speed Doesn't Mean Sacrificing Accuracy: The most critical insight is that the fast protocol's diagnostic reliability was not inferior to the standard protocol. This directly addresses the primary concern that accelerated imaging might lead to missed findings or misinterpretations.
  • Intra-Rater Reliability is Key: The finding that a single rater's consistency between the fast and clinical scans was higher than the consistency between different raters on the standard clinical scan is particularly compelling. This suggests the fast protocol might even offer a more consistent assessment experience.
  • DMT Eligibility Remains Robust: The protocol's ability to reliably assess eligibility for disease-modifying therapies is crucial for the practical implementation of new treatments.

Key Results

  • Time Savings: The fast dementia protocol reduced acquisition time by 63%, from an average of 17 minutes 39 seconds to 6 minutes 29 seconds. Each of the four core sequences in the fast protocol took less than 2 minutes.
  • Diagnostic Reliability: Inter-rater reliability coefficients (κ^G) for diagnosis were comparable between the fast (0.54) and clinical (0.58) scans, with a ratio of 0.92 (95% CI: 0.82-1.04), indicating non-inferiority.
  • Visual Scale Reliability: Reliability for visual scale ratings (MTA, Koedam, Fazekas) and microhemorrhage assessments was also comparable between scan types, with ratios close to one.
  • DMT Eligibility Reliability: The reliability for assessing radiological eligibility for amyloid-lowering DMTs was very similar between the fast (0.90) and clinical (0.88) scans, with a ratio of 1.02 (95% CI: 0.97-1.06).
  • Intra-Rater Consistency: Intra-rater reliability between the fast and clinical scans was consistently higher than the inter-rater reliability on the clinical scan, suggesting improved consistency for individual raters.

Limitations and Future Work

  • Single-Center, Single-System Study: The study was conducted at a single specialist center using a specific Siemens 3T system. This limits generalizability to other centers, scanner vendors, and field strengths.
  • Cohort Characteristics: The cohort may not fully represent the broader dementia population due to recruitment from a specialist clinic, potentially leading to younger participants and fewer common comorbidities or less severe cerebrovascular disease.
  • Neuroradiologist Focus: Scans were assessed by neuroradiologists, not general radiologists. While this ensures expert evaluation, it might not reflect the experience of a broader clinical audience.
  • Image Quality Not Formally Assessed: While the primary aim was diagnostic reliability, formal assessment of image quality and artifacts was not conducted. However, the authors note that any substantial degradation would likely have impacted reliability.
  • MRI Slot Duration: The study focused on core scan time, not the total MRI slot duration (including patient setup, scout scans, etc.). Optimizing these non-imaging steps is crucial as scan times decrease.
  • Limited Sequence Set: Only four key sequences were included. Patients requiring additional sequences (e.g., contrast-enhanced, diffusion-weighted) might still need longer scans.

Future Work suggested by the authors includes: * Optimizing fast protocols for different MRI systems and field strengths. * Exploring deep learning-based reconstruction methods to further improve speed and quality. * Investigating the impact of incorporating SWI into accelerated protocols for patient eligibility and ARIA monitoring. * Conducting multi-center evaluations to confirm generalizability. * Assessing the impact of accelerated protocols on total MRI slot duration.

Summary updated

Abstract

The fast dementia protocol with four core sequences reduced acquisition time by 63%. The fast scan showed non-inferior reliability for diagnosis and visual ratings. Assessment for disease-modifying therapy eligibility was similar between scan types. Fast protocols may improve access to magnetic resonance imaging and diagnosis in dementia.
BibTeX
@inproceedings{rosagrilo2025ultrafast,
  author = {Miguel Rosa‐Grilo and Haroon Chughtai and David L. Thomas and Daniel C. Alexander and Millie Beament and Christopher R S Belder and Hans Rolf Jäger and Emma Lim and Nicholas Magill and Dermot Mallon and Jennifer M. Nicholas and Owen Nicholas and F Urbán and Geoff J.M. Parker and Frederik Barkhof and Catherine J. Mummery and Nick C. Fox},
  title = {{Ultra‐fast MRI for dementia diagnosis and treatment eligibility: A prospective study}},
  booktitle = {Alzheimer},
  year = {2025},
  doi = {10.1002/alz.70341},
}
Figure