Gruia Dragos-Cristian, Giunchiglia Valentina, Braban Andra, Parkinson Niamh, Banerjee Soma, Kwan Joseph, Hellyer Peter J, Hampshire Adam, Geranmayeh Fatemeh
Department of Brain Sciences, Faculty of Medicine, Imperial College London, London, UK.
Imperial College Healthcare NHS Trust, London, UK.
EClinicalMedicine. 2025 Sep 2;88:103469. doi: 10.1016/j.eclinm.2025.103469. eCollection 2025 Oct.
Cognitive impairments are prevalent in many neurological disorders and remain underdiagnosed and poorly studied longitudinally. Unsupervised remote cognitive testing is an accessible, scalable, and cost-effective solution. However it often fails to separate cognitive deficits from commonly co-occurring motor impairments. To address this gap, we present a computational framework that isolates cognitive ability from motor impairment in self-administered digital tasks.
Stroke was chosen as a representative neurological disorder, as patients frequently experience both motor and cognitive impairments. Our validation analyses spanned 18 computerised tasks that were completed longitudinally by a cohort of stroke survivors (N = 171) collected as part of the IC3 study between 2022 and 2024, covering a broad spectrum of cognitive and motor domains across multiple timepoints within the first two years post-stroke. IC3 study was registered under NCT05885295 and IRAS:299333. The computational model was applied on trial-level data to disentangle the contribution of motor and cognitive processes. Bayesian Principal Component Analysis (PCA) was applied to the resultant data for dimensionality reduction purposes, while mixed effects regression models and multivariate canonical correlation analyses were used to assess the model's clinical utility.
In patients with motor hand impairment, standard accuracy performance metrics were confounded in 10 tasks (p < 0.05, FDR-corrected). In contrast, the Modelled Cognitive metrics obtained from the computational framework showed no significant effects of impaired hand (p > 0.05, FDR-corrected). Moreover, the Modelled Cognitive metrics correlated more strongly with clinical pen-and-paper scales (mean R = 0.65 vs 0.43) and functional outcomes (mean R = 0.16 vs 0.09). Brain-behaviour associations were stronger when using the Modelled Cognitive metrics, and revealed intuitive multivariate relationships with individual tasks.
We present converging evidence for the improved clinical utility and validity of the Modelled Cognitive metrics within neurological conditions characterised by co-occurring motor and cognitive deficits. Addressing the confounding effects of motor impairment improves the reliability and biological validity of self-administered digital assessments, potentially enhancing accessibility and supporting early detection and intervention across neurological disorders.
This research is funded by the UK Medical Research Council (MR/T001402/1). Infrastructure support was provided by the National Institute for Health Research (NIHR) Imperial Biomedical Research Centre and the NIHR Imperial Clinical Research Facility. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.
认知障碍在许多神经系统疾病中普遍存在，但仍未得到充分诊断，且纵向研究较少。无监督远程认知测试是一种可及、可扩展且具有成本效益的解决方案。然而，它往往无法将认知缺陷与常见的运动障碍区分开来。为了填补这一空白，我们提出了一个计算框架，该框架可在自我管理的数字任务中将认知能力与运动障碍隔离开来。
选择中风作为代表性的神经系统疾病，因为患者经常同时经历运动和认知障碍。我们的验证分析涵盖了18项计算机化任务，这些任务由一组中风幸存者（N = 171）在2022年至2024年期间作为IC3研究的一部分纵向完成，涵盖了中风后两年内多个时间点的广泛认知和运动领域。IC3研究已在NCT05885295和IRAS:299333注册。该计算模型应用于试验级数据，以区分运动和认知过程的贡献。贝叶斯主成分分析（PCA）应用于所得数据以进行降维，同时使用混合效应回归模型和多元典型相关分析来评估该模型的临床效用。
在有手部运动障碍的患者中，标准准确性性能指标在10项任务中受到混淆（p < 0.05，经FDR校正）。相比之下，从计算框架中获得的模拟认知指标显示手部受损无显著影响（p > 0.05，经FDR校正）。此外，模拟认知指标与临床纸笔量表（平均R = 0.65对0.43）和功能结果（平均R = 0.16对0.09）的相关性更强。使用模拟认知指标时，脑-行为关联更强，并揭示了与各个任务直观的多变量关系。
我们提供了越来越多的证据，证明在以同时存在运动和认知缺陷为特征的神经系统疾病中，模拟认知指标具有更高的临床效用和有效性。解决运动障碍的混杂效应提高了自我管理数字评估的可靠性和生物学有效性，有可能提高可及性，并支持对各种神经系统疾病的早期检测和干预。
本研究由英国医学研究理事会（MR/T001402/1）资助。基础设施支持由国家卫生研究院（NIHR）帝国生物医学研究中心和NIHR帝国临床研究设施提供。所表达的观点是作者的观点，不一定是NHS、NIHR或卫生与社会保健部的观点。
