Assessing verbal and spatial memory via smartphone

TL;DR

Researchers developed smartphone-based memory tests to detect cognitive decline in CNS diseases, finding that a spatial memory composite reliably tracks disability and brain damage.

Contribution

A novel smartphone-based spatial memory composite was developed and validated as a digital biomarker for mild cognitive dysfunction in CNS disorders.

Findings

The spatial memory composite distinguished healthy donors from various CNS disorders with high statistical significance.

The composite strongly correlated with disability scales and brain lesion volume when adjusted for sensory-motor delays.

A linear model using spatial memory biomarkers accurately predicted cognitive test scores in an independent cohort.

Abstract

Detecting subtle cognitive decline in chronic central nervous system (CNS) disease is hampered by practice effects, motor confounds, and the lack of premorbid baselines. Smartphone testing offers frequent unsupervised assessments but requires rigorous validation and internal quality control. Therefore, we aimed to develop and validate smartphone-based verbal and spatial memory tasks, derive composite digital biomarkers, and determine their utility for monitoring cognitive impairment.

In a prospective study (2018–2025) we enrolled 315 adults [41 healthy donors, 217 people with Multiple Sclerosis (MS), 57 other CNS disorders]. Participants completed 5,875 verbal and 2,588 spatial trials on the Neurological Functions Test Suite App. We extracted fourteen digital biomarkers reflecting individualized Sensory-Motor Processing Thresholds (iSMPT), adjusted test latencies and test accuracies, and split the MS cohort into training (n = 145) and validation (n = 72) sets. Ordinal logistic regression generated composite scores that were tested in the held-out cohort. Reliability (intraclass correlation coefficient, ICC) and criterion validity (Spearman Rho, R², Lin's concordance) were calculated; p-values < 0.05 were considered significant.

The spatial memory composite separated healthy donors from relapsing-remitting multiple sclerosis (RR-MS), progressive MS, non-inflammatory and inflammatory neurological diseases, and radiologically/clinically isolated syndromes (all P < 0.05), outperforming all single biomarkers. With iSMPT included, it correlated most strongly with global disability [Expanded Disability Status scale [EDSS], Neurological exam composite [NeurEx]; Rho = 0.62–0.63, P < 0.001] and lesion volume (Rho = 0.71, P = 0.008); without iSMPT it tracked cognitive subscores [EDSS Functional System Score 7 (FSS7), NeurEx panel 1; Rho = 0.50–0.53, P < 0.001] and Combinatorial MRI Score of CNS tissue destruction (COMRIS-CTD; Rho = 0.63, P < 0.001). Test–retest reliability was good (ICC = 0.813 with iSMPT; 0.713 without). A linear model using spatial memory biomarkers predicted Symbol Digit Modalities Test (SDMT) scores in an independent cohort (n = 63; R² = 0.53, Rho = 0.74, concordance = 0.66), enabling a 95% prediction envelope for real-time quality control.

The spatial memory composite provides a rapid, reliable, and sensory-motor delay-adjusted digital marker of mild cognitive dysfunction. Coupled with SDMT, it enhances unsupervised monitoring and internal quality assurance across diverse CNS disorders.

Figures