Gait Analysis Using Wearable Sensors in Patients With Alzheimer's Dementia: A Preliminary Report.

Stroke is a leading cause of disability and death, with an estimated 12 million strokes occurring annually worldwide []. Acute stroke diagnosis and treatment rely on clinical assessment (i.e. motor, facial, sensory, and language deficits) and imaging to determine aetiology and lesion location [, , ]. There is, however, increased recognition of the high prevalence of cognitive impairment at all stages after stroke []. Post-stroke cognitive impairment (PSCI) is defined as a cognitive deficit that develops in the first three months following stroke, persists for a minimum of six months, cannot be explained by another condition, and involves language, executive function, visuospatial, episodic, or working memory cognitive domains [,]. PSCI is highly relevant to stroke recovery and portends a significant dementia risk, as reflected by 59% of patients with cognitive impairment at three months[], and approximately 33% of patients with PSCI that convert to dementia in the years post-stroke [,,]. Research on screening to PSCI is warranted, particularly exploiting imaging where biomarkers of ageing and cognition are well established. Indeed, comprehensive cognitive work-up in the early stages after stroke and during recovery is resource intensive. It requires dedicated personnel and may burden patients and their caregivers [,]. Non-contrast Computed Tomography (CT) imaging is ubiquitous in the acute stroke workup. CT images contain morphological and anatomical details that may relate to cognition, in addition to the stroke related imaging information [, , ]. For instance, the brain’s ventricles are easily identifiable on CT [], providing the means for segmentation. Although ventricle volume is not reported as part of standard acute stroke care, ventricular enlargement and/or brain tissue atrophy [,] are independent markers of brain health, cognitive decline, and long-term outcomes after stroke [,,]. Quantification of the lateral ventricle volume (LVV) is not currently used to inform on PSCI risk despite an association that relates to neurodegeneration processes []. Hence there is potential clinical value in fully automated segmentation methods to estimate ventricle volumes on CT []. The current study focuses on LVV in relation to cognitive, functional, and stroke severity outcomes. The Norwegian COgnitive impairment After STroke (Nor-COAST) study offers data that are uniquely suited for the current investigation, as it involved baseline imaging and up to three years of longitudinal assessments among stroke survivors []. The primary objective was to test for associations between LVV assessed at acute presentation and the Montreal Cognitive Assessment (MoCA) score over time. The study also aims to investigate cross-sectional relationships between the LVV and other cognitive, clinical, and stroke impairment scales. To facilitate the LVV estimate we deployed an automated deep learning tool that segments the lateral ventricles, along with any type of intracranial hemorrhage (called ventricle and bleed artificial intelligence (AI), VB-AI), which was designed for use in both stroke and traumatic brain injury [].