Investigating neural markers of Alzheimer's disease in posttraumatic stress disorder using machine learning algorithms and magnetic resonance imaging

Gabriella Yakemow; Tiffany A Kolesar; Natalie Wright; Iman Beheshti; Eun Hyung Choi; Lawrence Ryner; Sarah Chaulk; Ronak Patel; Ji Hyun Ko

doi:10.3389/fneur.2024.1470727

Investigating neural markers of Alzheimer's disease in posttraumatic stress disorder using machine learning algorithms and magnetic resonance imaging

Front Neurol. 2024 Nov 7:15:1470727. doi: 10.3389/fneur.2024.1470727. eCollection 2024.

Authors

Gabriella Yakemow^{1

2}, Tiffany A Kolesar^{1

2}, Natalie Wright^{1

2

3}, Iman Beheshti^{1

2}, Eun Hyung Choi^{1

2}, Lawrence Ryner⁴, Sarah Chaulk⁵, Ronak Patel⁵, Ji Hyun Ko^{1

2

6}

Affiliations

¹ Department of Human Anatomy and Cell Science, University of Manitoba, Winnipeg, MB, Canada.
² PrairieNeuro Brain Research Centre, Health Sciences Centre, Kleysen Institute for Advanced Medicine, Winnipeg, MB, Canada.
³ Undergraduate Medical Education, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB, Canada.
⁴ Department of Radiology, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB, Canada.
⁵ Department of Clinical Health Psychology, University of Manitoba, Winnipeg, MB, Canada.
⁶ Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB, Canada.

Abstract

Introduction: Posttraumatic stress disorder (PTSD) is a mental health disorder caused by experiencing or witnessing traumatic events. Recent studies show that patients with PTSD have an increased risk of developing dementia, including Alzheimer's disease (AD), but there is currently no way to predict which patients will go on to develop AD. The objective of this study was to identify structural and functional neural changes in patients with PTSD that may contribute to the future development of AD.

Methods: Neuroimaging (pseudo-continuous arterial spin labeling [pCASL] and structural magnetic resonance imaging [MRI]) and behavioral data for the current study (n = 67) were taken from our non-randomized open label clinical trial (ClinicalTrials.gov Identifier: NCT03229915) for treatment-seeking individuals with PTSD (n = 40) and age-matched healthy controls (HC; n = 27). Only the baseline measures were utilized for this study. Mean cerebral blood flow (CBF) and gray matter (GM) volume were compared between groups. Additionally, we utilized two previously established machine learning-based algorithms, one representing AD-like brain activity (Machine learning-based AD Designation [MAD]) and the other focused on AD-like brain structural changes (AD-like Brain Structure [ABS]). MAD scores were calculated from pCASL data and ABS scores were calculated from structural T₁-MRI images. Correlations between neuroimaging data (regional CBF, GM volume, MAD scores, ABS scores) and PTSD symptom severity scores measured by the clinician-administered PTSD scale for DSM-5 (CAPS-5) were assessed.

Results: Decreased CBF was observed in two brain regions (left caudate/striatum and left inferior parietal lobule/middle temporal lobe) in the PTSD group, compared to the HC group. Decreased GM volume was also observed in the PTSD group in the right temporal lobe (parahippocampal gyrus, middle temporal lobe), compared to the HC group. GM volume within the right temporal lobe cluster negatively correlated with CAPS-5 scores and MAD scores in the PTSD group.

Conclusion: Results suggest that patients with PTSD with reduced GM volume in the right temporal regions (parahippocampal gyrus) experienced greater symptom severity and showed more AD-like brain activity. These results show potential for early identification of those who may be at an increased risk for future development of dementia.

Keywords: Alzheimer's disease; MRI; arterial spin labeling (ASL); machine learning; posttraumatic stress disorder (PTSD).

Associated data

ClinicalTrials.gov/NCT03229915

Grants and funding

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. Funding support was provided by the Workers Compensation Board of Manitoba (grant number 16-03), the Natural Sciences and Engineering Research Council of Canada (RGPIN-2016-05964; RGPIN-2023-04283; PDF-2020-545615), and the University of Manitoba.