Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Forecasting Future Anatomies: Longitudinal Brain Mri-to-Mri Prediction

Ali Farki, Elaheh Moradi, Deepika Koundal, Jussi Tohka

TL;DR

This study tackles the problem of forecasting an individual’s future brain anatomy from a single baseline MRI by treating it as image-to-image longitudinal prediction. It compares five deep learning architectures (UNet, U$^2$-Net, UNETR, TEUNet, and ODE-UNet) across two cohorts (ADNI and AIBL) using voxel-level gray matter density maps and evaluates both global reconstruction metrics ($MSE$, $PSNR$, $SSIM$) and longitudinal change with $\

Abstract

Predicting future brain state from a baseline magnetic resonance image (MRI) is a central challenge in neuroimaging and has important implications for studying neurodegenerative diseases such as Alzheimer's disease (AD). Most existing approaches predict future cognitive scores or clinical outcomes, such as conversion from mild cognitive impairment to dementia. Instead, here we investigate longitudinal MRI image-to-image prediction that forecasts a participant's entire brain MRI several years into the future, intrinsically modeling complex, spatially distributed neurodegenerative patterns. We implement and evaluate five deep learning architectures (UNet, U2-Net, UNETR, Time-Embedding UNet, and ODE-UNet) on two longitudinal cohorts (ADNI and AIBL). Predicted follow-up MRIs are directly compared with the actual follow-up scans using metrics that capture global similarity and local differences. The best performing models achieve high-fidelity predictions, and all models generalize well to an independent external dataset, demonstrating robust cross-cohort performance. Our results indicate that deep learning can reliably predict participant-specific brain MRI at the voxel level, offering new opportunities for individualized prognosis.

Forecasting Future Anatomies: Longitudinal Brain Mri-to-Mri Prediction

TL;DR

This study tackles the problem of forecasting an individual’s future brain anatomy from a single baseline MRI by treating it as image-to-image longitudinal prediction. It compares five deep learning architectures (UNet, U-Net, UNETR, TEUNet, and ODE-UNet) across two cohorts (ADNI and AIBL) using voxel-level gray matter density maps and evaluates both global reconstruction metrics (, , ) and longitudinal change with $\

Abstract

Predicting future brain state from a baseline magnetic resonance image (MRI) is a central challenge in neuroimaging and has important implications for studying neurodegenerative diseases such as Alzheimer's disease (AD). Most existing approaches predict future cognitive scores or clinical outcomes, such as conversion from mild cognitive impairment to dementia. Instead, here we investigate longitudinal MRI image-to-image prediction that forecasts a participant's entire brain MRI several years into the future, intrinsically modeling complex, spatially distributed neurodegenerative patterns. We implement and evaluate five deep learning architectures (UNet, U2-Net, UNETR, Time-Embedding UNet, and ODE-UNet) on two longitudinal cohorts (ADNI and AIBL). Predicted follow-up MRIs are directly compared with the actual follow-up scans using metrics that capture global similarity and local differences. The best performing models achieve high-fidelity predictions, and all models generalize well to an independent external dataset, demonstrating robust cross-cohort performance. Our results indicate that deep learning can reliably predict participant-specific brain MRI at the voxel level, offering new opportunities for individualized prognosis.

Paper Structure

This paper contains 9 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Datasets and Preprocessing
  4. Deep Learning Architectures, Training and Evaluation
  5. Results
  6. ADNI Test Set Performance
  7. AIBL External Validation
  8. Error Analysis
  9. Discussion

Figures (4)

  • Figure 1: MSE comparison between ADNI and AIBL test sets.
  • Figure 2: Voxel-wise MSE maps averaged across ADNI test participants. Axial slices at $z = 14mm$ of the MNI space are shown.
  • Figure 3: Histogram of voxel-wise correlations between the predicted 24-month image and the actual 24-month image across participants in ADNI test set (ODE-UNet trained on BigDataset).
  • Figure 4: Histogram of voxel-wise $\Delta$-Pearson correlations on ADNI test set (ODE-UNet trained on BigDataset).