Testing a convolutional neural network‐based hippocampal segmentation method in a stroke population

Abstract

As stroke mortality rates decrease, there has been a surge of effort to study poststroke dementia (PSD) to improve long-term quality of life for stroke survivors. Hippocampal volume may be an important neuroimaging biomarker in poststroke dementia, as it has been associated with many other forms of dementia. However, studying hippocampal volume using MRI requires hippocampal segmentation.Advances in automated segmentation methods have allowed for studying the hippocampus on a large scale, which is important for robust results in the heterogeneous stroke population. However, most of these automated methods use a single atlasbased approach and may fail in the presence of severe structural abnormalities common in stroke. Hippodeep, a new convolutional neural network-based hippocampal segmentation method, does not rely solely on a single atlas-based approach and thus may be better suited for stroke populations. Here, we compared quality control and the accuracy of segmentations generated by Hippodeep and two well-accepted hippocampal segmentation methods on stroke MRIs (FreeSurfer 6.0 whole hippocampus and FreeSurfer 6.0 sum of hippocampal subfields). Quality control was performed using a stringent protocol for visual inspection of the segmentations, and accuracy was measured as volumetric correlation with manual segmentations. Hippodeep performed significantly better than both FreeSurfer methods in terms of quality control.All three automated segmentation methods had good correlation with manual segmentations and no one method was significantly more correlated than the others.Overall, this study suggests that both Hippodeep and FreeSurfer may be useful for hippocampal segmentation in stroke rehabilitation research, but Hippodeep may be more robust to stroke lesion anatomy.

Cortical Thickness Estimation in Individuals With Cerebral Small Vessel Disease, Focal Atrophy, and Chronic Stroke Lesions

“…While small acute strokes may have minimal effects on tissue segmentation, large chronic cortico-subcortical stroke lesions introduce alterations to brain morphometry resulting in failed segmentation in most brain segmentation algorithms ( Wang et al, 2012 ; Yang et al, 2016 ; Siegel et al, 2017 ; Zavaliangos-Petropulu et al, 2020 ). Although this issue is particularly relevant in individuals with CVD, cerebral small vessel disease and brain atrophy that are commonly observed in patients with Alzheimer’s and other related dementias present similar challenges when estimating cortical thickness.…”

Chronic Stroke Sensorimotor Impairment Is Related to Smaller Hippocampal Volumes: An ENIGMA Analysis

“… 27 Hippodeep was previously found to be the most robust out of the freely available methods for segmenting the hippocampus in people with stroke pathology. 28 Hippocampal segmentations were visually inspected according to previously described protocols. 23 , 28 Any segmentations that were not properly segmented were marked as failed and excluded from the analysis.…”

“… 28 Hippocampal segmentations were visually inspected according to previously described protocols. 23 , 28 Any segmentations that were not properly segmented were marked as failed and excluded from the analysis. This resulted in different sample sizes for the ipsilesional and contralesional analyses.…”

The Enhancing NeuroImaging Genetics through Meta‐Analysis Consortium: 10 Years of Global Collaborations in Human Brain Mapping

“…Complementary approaches that use deep learning for anatomical segmentation of the hippocampus are compared and reviewed in an empirical study by Zavaliangos-Petropulu et al (2020). With the renewed interest in deep-learning methods and their application to largescale biobanks, we may soon be able to automate the quality control of data from many more imaging modalities (Petrov et al, 2018).…”