Ultra-fast MRI for brain-age prediction in a real-world cognitive disorders clinic

Apr 3, 2026Frontiers in aging neuroscience

Using very fast MRI to estimate brain age in a clinic for thinking problems

AI simplified

Abstract

Excellent cross-protocol agreement (intraclass correlation coefficient: ICC ≳ 0.90) was observed for brain-age estimates between standard and Wave-CAIPI MRI scans.

  • Clinical discrimination between subjective memory complaints and neurodegenerative disorders was comparable across both MRI protocols.
  • Small, model-specific offsets and significant interactions between acquisition method and diagnosis were noted for some brain-age pipelines.
  • Test-retest reliability was high, indicating consistency in brain-age measurements.
  • Quality control measures were similar across both standard and ultra-fast scan protocols.

AI simplified

Key numbers

0.90
Intraclass Correlation Coefficient (ICC)
Cross-protocol agreement across brain-age models
3.73 years
Mean Absolute Error (MAE)
Accuracy of brain-age prediction for standard MPRAGE
2 years
Brain-age gap increase
Difference in brain-age gap across patient groups

Full Text

What this is

  • This research investigates the use of ultra-fast MRI (Wave-CAIPI) for predicting brain age in patients with cognitive disorders.
  • It compares the performance of this new imaging protocol against the standard MRI method (MPRAGE) in a clinical setting.
  • The study assesses the accuracy and reliability of brain-age estimates across different software models, focusing on their clinical applicability.

Essence

  • Ultra-fast Wave-CAIPI MRI provides reliable brain-age estimates comparable to standard MPRAGE in cognitive disorder patients, reducing scan time significantly. The study demonstrates that brain-age models can effectively differentiate between subjective memory complaints and neurodegenerative disorders, although some model-specific biases exist.

Key takeaways

  • Wave-CAIPI MRI achieves high cross-protocol agreement with brain-age models, with intraclass correlation coefficients (ICC) exceeding 0.90. This indicates that the ultra-fast scans can reliably produce brain-age estimates similar to those derived from standard MRI.
  • The model pyment showed the highest prediction accuracy for brain age, with a mean absolute error (MAE) of 3.73 years for standard MPRAGE and 4.96 years for Wave-CAIPI. This performance suggests that pyment is particularly effective in clinical settings.
  • Patients with neurodegenerative disorders exhibited a larger brain-age gap compared to those with subjective memory complaints, with differences of approximately 2 years depending on the imaging protocol. This underscores the potential of MRI-derived brain-age estimates in clinical diagnosis.

Caveats

  • Some brain-age models demonstrated protocol-related biases, with differences in predicted brain age ranging from -3.12 years to +1.55 years depending on the model. This variability suggests that harmonization techniques may be necessary when integrating different MRI protocols.
  • The study's sample size for the short-interval test-retest analysis was limited to 15 participants, which may affect the robustness of the findings regarding within-subject reliability.

AI simplified

what lands in your inbox each week:

  • 📚7 fresh studies
  • 📝plain-language summaries
  • direct links to original studies
  • 🏅top journal indicators
  • 📅weekly delivery
  • 🧘‍♂️always free