Combining PropSeg and a convolutional neural network for automatic spinal cord segmentation in pediatric populations and patients with spinal cord injury

Abstract

Segmentation of the spinal cord is an essential process for the accurate delineation of spinal cord structures but can be a tedious task for experts when using manual or semi‐automated tools. On the other hand, existing automatic segmentation algorithms have not been developed with the pediatric or injured spinal cord in mind. This study presents a novel automated segmentation method that combines the flexibility of deterministic approaches and the powerfulness of neural networks, applied to pediatric and injured spinal cord magnetic resonance imaging (MRI) data. The method first applies the PropSeg algorithm several times on small patches of the spinal cord MRI with various initialization parameters. Then, a convolutional neural network concatenates all these small segmentations with the original MR images to compute a final segmentation. Our results demonstrate good performances on the whole spinal cord (Dice score = 0.88 vs. 0.9) while outperforming existing methods on spinal cord injury regions (0.8 vs. 0.63).

SCIseg: Automatic Segmentation of T2-weighted Intramedullary Lesions in Spinal Cord Injury

“…Similarly, as the injury levels in the training dataset were skewed towards the cervical spine, its ability to segment lumbar lesions is expected to be lower compared to cervical/thoracic lesions. [R3.2, R3.8] Only a few studies exist in the literature discussing the importance of automatic segmentation in SCI scans (24,36). McCoy et al's study (24) is the closest to ours as it presented the first DL method for the segmentation of SC and intramedullary lesions in SCI.…”

“…As a result, researchers often (semi) manually annotate both spinal cord and lesions in SCI patients (6,13,14), which is a time-consuming process susceptible to intra- and inter-rater variability (2). Only a few studies exist in the literature discussing the importance of automatic segmentation in SCI scans (20,31). McCoy et al’s study (20) is the closest to ours as it presented the first DL method for the segmentation of SC and lesions in SCI.…”

“…Only a few studies exist in the literature discussing the importance of automatic segmentation in SCI scans (24,36). McCoy et al’s study (24) is the closest to ours as it presented the first DL method for the segmentation of SC and intramedullary lesions in SCI.…”

Reproducible Spinal Cord Quantitative MRI Analysis with the Spinal Cord Toolbox

“…It has rapidly become a reference dataset for methods development. 35 – 42 The Spinal Cord Head Position MRI dataset ( ) contains T2w scans from 10 healthy participants with 3 different neck positions: flexion, neutral, and extension. The dataset was acquired in the context of evaluating the pontomedullary junction as a reference for computing the spinal cord cross‑sectional area.…”

“…Current research is focused on the development of additional segmentation tools, such as contrast-agnostic spinal cord segmentation, 65 spinal cord segmentation in pediatric populations, 42 segmentation of multiple sclerosis lesions across hospitals using continual learning, 66 multi-site segmentation of spinal cord injury lesions, 67 segmentation of spinal nerve rootlets, 68 and deep learning-based intervertebral disc labeling. 35 …”