Accurate, rapid and reliable, fully automated MRI brainstem segmentation for application in multiple sclerosis and neurodegenerative diseases

Abstract

Neurodegenerative disorders, such as Alzheimer's disease (AD) and progressive forms of multiple sclerosis (MS), can affect the brainstem and are associated with atrophy that can be visualized by MRI. Anatomically accurate, large-scale assessments of brainstem atrophy are challenging due to lack of automated, accurate segmentation methods. We present a novel method for brainstem volumetry using a fully-automated segmentation approach based on multi-dimensional gated recurrent units (MD-GRU), a deep learning based semantic segmentation approach employing a convolutional adaptation of gated recurrent units. The neural network was trained on 67 3D-high resolution T1-weighted MRI scans from MS patients and healthy controls (HC) and refined using segmentations of 20 independent MS patients' scans. Reproducibility was assessed in MR test-retest experiments in 33 HC. Accuracy and robustness were examined by Dice scores comparing MD-GRU to FreeSurfer and manual brainstem segmentations in independent MS and AD datasets. The mean %-change/SD between test-retest brainstem volumes were 0.45%/0.005 (MD-GRU), 0.95%/0.009 (FreeSurfer), 0.86%/0.007 (manually edited segmentations). Comparing MD-GRU to manually edited segmentations the mean Dice scores/SD were: 0.97/0.005 (brainstem), 0.95/0.013 (mesencephalon), 0.98/0.006 (pons), 0.95/0.015 (medulla oblongata). Compared to the manual gold standard, MD-GRU brainstem segmentations were more accurate than FreeSurfer segmentations (p < .001). In the multi-centric acquired AD data, the mean Dice score/SD for the MD-GRU-manual segmentation comparison was 0.97/0.006. The fully automated brainstem segmentation method MD-GRU provides accurate, highly reproducible, and robust segmentations in HC and patients with MS and AD in 200 s/scan on an Nvidia GeForce GTX 1080 GPU and shows potential for application in large and longitudinal datasets.

Keywords: MD-GRU; atrophy; brainstem; deep learning; multiple sclerosis; segmentation.

Figure 1

Schematic figure of (a) training and refinement and (b) accuracy, reproducibility and robustness assessment of the algorithm [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2

Illustrations of anatomical landmarks used in the manual segmentation of the mesencephalon, pons and medulla oblongata. Caudal delimitation of the MO is defined as the bilateral exit of the first spinal root (white arrows) in axial slices. The pontomedullary sulcus marks the cranial delimitation (white arrowheads). The pontomesencephalic junction is marked by the black arrow, the cranial delimitation of the mesencephalon toward the pineal gland is shown by the black arrowhead

Figure 3A

Association between manually edited segmentations and MD‐GRU segmentations for the total brainstem and its substructures mesencephalon, pons and medulla oblongata (n = 80) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3B

Bland–Altman plots showing the absolute differences between manually edited and MD‐GRU based segmentations plotted against their averages. The dashed lines indicate the limits of agreement (mean+/−1.96 SD), (n = 80)

Figure 4A

Comparison of MD‐GRU and manually edited segmentations of 30 exemplary subjects in the accuracy data set (n = 80) and corresponding Dice scores for total brainstem volumes [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 4B

Exemplary axial and coronal views of the brainstem MD‐GRU segmentations [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 5

Example of random imprecise segmentations of the segmentation algorithm MD‐GRU. (a) Missing voxels in the ponto‐mesencephalic junction. (b) Segmented voxels in the pineal gland

Figure 6

MD‐GRU segmentations of the eleven 1.5 T scan‐rescan experiments and corresponding mean percentage changes for total BS volumes. Please note one outlier with 1.62% BS volume change between the two scans, associated with a pronounced anteversion of the brainstem axis of this subject compared to the other subjects