Improving Accuracy of Brainstem MRI Volumetry: Effects of Age and Sex, and Normalization Strategies

Laura Sander, Antal Horváth, Simon Pezold, Simon Andermatt, Michael Amann, Tim Sinnecker, Maria Janina Wendebourg, Eva Kesenheimer, Özgür Yaldizli, Ludwig Kappos, Cristina Granziera, Jens Wuerfel, Philippe C. Cattin, Regina SchlaegerFrontiers in Neuroscience 2020 · DOI

Abstract

Brainstem-mediated functions are often impaired in neurodegenerative diseases and aging. This study investigates extrinsic and intrinsic sources of brainstem volume variability and develops normalization strategies to improve the accuracy of MRI-based volumetry. Using a deep-learning segmentation algorithm (MD-GRU), researchers found that normalizing for total intracranial volume (TICV) reduces inter-individual variability by 35%, while the brainstem itself remains remarkably resilient to aging in healthy subjects.

Introduction

The brainstem is a vital "control center" for life-sustaining functions, including respiration and heart rate. Atrophy in this region is a hallmark of several neurodegenerative diseases, such as Multiple Sclerosis (MS), Parkinson’s, and Alzheimer’s. However, clinical use of brainstem volume as a biomarker is complicated by "noise" from two primary sources: extrinsic variability (differences in MRI scanners and protocols) and intrinsic variability (natural anatomical differences between individuals). This study aims to quantify these factors and propose efficient normalization models to enhance the sensitivity and reliability of brainstem measurements.

Methods

The researchers employed the MD-GRU (Multi-dimensional Gated Recurrent Units) deep-learning algorithm to automatically segment the brainstem into its three primary substructures: the mesencephalon, pons, and medulla oblongata.

MRI composite showing brainstem segmentation (mesencephalon in teal, pons in green, medulla in blue) and anatomical landmarks used for normalization.

The study utilized three distinct groups of healthy subjects (HS):

  1. Scanner Comparison Group: 11 HS scanned on different 3T machines to assess hardware-related variability.
  2. Protocol Comparison Group: 22 HS scanned with varying imaging parameters on a 1.5T machine.
  3. Variability/Normalization Group: 110 HS (ages 18–72) used to assess anatomical variability and build normalization models.

Normalization models were constructed using statistical "backward selection" to identify the most significant anatomical covariates, such as total intracranial volume (TICV), age, and sex.

Results

The MD-GRU algorithm demonstrated remarkable robustness across different imaging environments, with Dice scores (a measure of segmentation overlap) consistently above 0.94. This suggests the tool is "scanner-agnostic" and well-suited for multi-center clinical trials.

Key findings include:

  • Variability Reduction: Normalizing brainstem volumes by TICV or a volumetric scaling factor (v-scale) reduced the relative standard deviation of measurements by approximately 35%. This significantly tightens the "normal" range, allowing for more precise detection of pathological changes.
  • Age Resilience: Unlike the cerebral cortex, which typically shrinks with age, the brainstem showed no significant volume reduction in this healthy cohort up to age 72. This makes any observed shrinkage a much stronger "red flag" for disease.
  • Sex and the Medulla: Most volume differences between men and women were explained by head size (TICV). However, the medulla oblongata remained significantly larger in men even after normalization, a biological finding that remains an open question for future research.

Discussion and Conclusion

The study confirms that deep-learning-based brainstem segmentation is highly consistent across different scanners and protocols. The substantial natural variation in brainstem volume (approx. 9.8%) can be effectively mitigated through TICV-based normalization. Because the brainstem appears resilient to normal aging, these refined volumetric measurements serve as highly sensitive biomarkers for detecting pathological atrophy. Future work will involve validating these models in patient groups, such as those with Progressive Supranuclear Palsy, to improve diagnostic accuracy in clinical settings.

Summary updated

Abstract

Background: Brainstem-mediated functions are impaired in neurodegenerative diseases and aging. Atrophy can be visualized by MRI. This study investigates extrinsic sources of brainstem volume variability, intrinsic sources of anatomical variability, and the influence of age and sex on the brainstem volumes in healthy subjects. We aimed to develop efficient normalization strategies to reduce the effects of intrinsic anatomic variability on brainstem volumetry. Methods: Brainstem segmentation was performed from MPRAGE data using our deep-learning-based brainstem segmentation algorithm MD-GRU. The extrinsic variability of brainstem volume assessments across scanners and protocols was investigated in two groups comprising 11 (median age 33.3 years, 7 women) and 22 healthy subjects (median age 27.6 years, 50% women) scanned twice and compared using Dice scores. Intrinsic anatomical inter-individual variability and age and sex effects on brainstem volumes were assessed in segmentations of 110 healthy subjects (median age 30.9 years, range 18–72 years, 53.6% women) acquired on 1.5T (45%) and 3T (55%) scanners. The association between brainstem volumes and predefined anatomical covariates was studied using Pearson correlations. Anatomical variables with associations of | r | > 0.30 as well as the variables age and sex were used to construct normalization models using backward selection. The effect of the resulting normalization models was assessed by % relative standard deviation reduction and by comparing the inter-individual variability of the normalized brainstem volumes to the non-normalized values using paired t- tests with Bonferroni correction. Results: The extrinsic variability of brainstem volumetry across different field strengths and imaging protocols was low (Dice scores > 0.94). Mean inter-individual variability/SD of total brainstem volumes was 9.8%/7.36. A normalization based on either total intracranial volume (TICV), TICV and age, or v-scale significantly reduced the inter-individual variability of total brainstem volumes compared to non-normalized volumes and similarly reduced the relative standard deviation by about 35%. Conclusion: The extrinsic variability of the novel brainstem segmentation method MD-GRU across different scanners and imaging protocols is very low. Anatomic inter-individual variability of brainstem volumes is substantial. This study presents efficient normalization models for variability reduction in brainstem volumetry in healthy subjects.
BibTeX
@inproceedings{sander2020improving,
  author = {Laura Sander and Antal Horváth and Simon Pezold and Simon Andermatt and Michael Amann and Tim Sinnecker and Maria Janina Wendebourg and Eva Kesenheimer and Özgür Yaldizli and Ludwig Kappos and Cristina Granziera and Jens Wuerfel and Philippe C. Cattin and Regina Schlaeger},
  title = {{Improving Accuracy of Brainstem MRI Volumetry: Effects of Age and Sex, and Normalization Strategies}},
  booktitle = {Frontiers in Neuroscience},
  year = {2020},
  doi = {10.3389/fnins.2020.609422},
}
Figure