Title

Leave-One-Out Cross-Validation and Linear Modeling of Visuospatial Memory to Predict Long-Term Motor Skill Retention in Individuals With and Without Chronic Stroke: A Short Report

Document Type

Article

Publication Date

10-15-2020

Publication Title

bioRxiv

First page number:

1

Last page number:

31

Abstract

Motor learning is fundamental to motor rehabilitation outcomes and has been associated with visuospatial memory function in previous studies. Current predictive models of motor recovery of individuals with stroke generally exclude cognitive measures, overlooking the connection between motor learning and visuospatial memory. Recent work has demonstrated that a clinical test of visuospatial memory (Rey-Osterrieth Complex Figure Delayed Recall) may predict one-month skill learning in older adults, but if this relationship persists in individuals with chronic stroke remains unknown. The purpose of this short report was to extend these findings by evaluating the extent these test scores impacted prediction in older adults and determine if this relationship generalized to individuals with stroke pathology. To address these questions, we trained two regression models (one including Delayed Recall scores and one without) using data from non-stroke older adults. To determine the extent to which Delayed Recall test scores impacted prediction accuracy of one-month skill learning in older adults, we used leave-one-out cross-validation to evaluate the prediction error between models. To determine if this predictive relationship persisted in individuals with chronic ischemic stroke, we then tested each trained model on an independent stroke dataset. Results indicated that in both stroke and non-stroke datasets, inclusion of Delayed Recall scores explained significantly more variance of one-month skill performance than models that included age, education, and baseline motor performance alone. This proof-of-concept suggests that the relationship between delayed visuospatial memory and one-month motor skill performance generalizes to individuals with chronic stroke and supports the idea that visuospatial testing may provide prognostic insight into motor rehabilitation outcomes.

Keywords

Model validation; Motor learning; Stroke rehabilitation; Upper extremity

Disciplines

Kinesiology | Life Sciences | Motor Control

Language

English

UNLV article access

Search your library

Share

COinS