Detecting Motor Impairment in Early Parkinson's Disease via Natural Typing Interaction With Keyboards: Validation of the neuroQWERTY Approach in an Uncontrolled At-Home Setting - PubMed (original) (raw)
Detecting Motor Impairment in Early Parkinson's Disease via Natural Typing Interaction With Keyboards: Validation of the neuroQWERTY Approach in an Uncontrolled At-Home Setting
Teresa Arroyo-Gallego et al. J Med Internet Res. 2018.
Abstract
Background: Parkinson's disease (PD) is the second most prevalent neurodegenerative disease and one of the most common forms of movement disorder. Although there is no known cure for PD, existing therapies can provide effective symptomatic relief. However, optimal titration is crucial to avoid adverse effects. Today, decision making for PD management is challenging because it relies on subjective clinical evaluations that require a visit to the clinic. This challenge has motivated recent research initiatives to develop tools that can be used by nonspecialists to assess psychomotor impairment. Among these emerging solutions, we recently reported the neuroQWERTY index, a new digital marker able to detect motor impairment in an early PD cohort through the analysis of the key press and release timing data collected during a controlled in-clinic typing task.
Objective: The aim of this study was to extend the in-clinic implementation to an at-home implementation by validating the applicability of the neuroQWERTY approach in an uncontrolled at-home setting, using the typing data from subjects' natural interaction with their laptop to enable remote and unobtrusive assessment of PD signs.
Methods: We implemented the data-collection platform and software to enable access and storage of the typing data generated by users while using their computer at home. We recruited a total of 60 participants; of these participants 52 (25 people with Parkinson's and 27 healthy controls) provided enough data to complete the analysis. Finally, to evaluate whether our in-clinic-built algorithm could be used in an uncontrolled at-home setting, we compared its performance on the data collected during the controlled typing task in the clinic and the results of our method using the data passively collected at home.
Results: Despite the randomness and sparsity introduced by the uncontrolled setting, our algorithm performed nearly as well in the at-home data (area under the receiver operating characteristic curve [AUC] of 0.76 and sensitivity/specificity of 0.73/0.69) as it did when used to evaluate the in-clinic data (AUC 0.83 and sensitivity/specificity of 0.77/0.72). Moreover, the keystroke metrics presented a strong correlation between the 2 typing settings, which suggests a minimal influence of the in-clinic typing task in users' normal typing.
Conclusions: The finding that an algorithm trained on data from an in-clinic setting has comparable performance with that tested on data collected through naturalistic at-home computer use reinforces the hypothesis that subtle differences in motor function can be detected from typing behavior. This work represents another step toward an objective, user-convenient, and quasi-continuous monitoring tool for PD.
Keywords: eHealth; machine learning; telemedicine.
©Teresa Arroyo-Gallego, María J Ledesma-Carbayo, Ian Butterworth, Michele Matarazzo, Paloma Montero-Escribano, Verónica Puertas-Martín, Martha L Gray, Luca Giancardo, Álvaro Sánchez-Ferro. Originally published in the Journal of Medical Internet Research (http://www.jmir.org), 26.03.2018.
Conflict of interest statement
Conflicts of Interest: TAG is currently employed by nQ Medical Inc, a company that has licensed the neuroQWERTY technology.
Figures
Figure 1
At-home typing activity. Panel A represents the amount of typing data collected from each of the 52 subjects (25 PwP, 27 CNT) included in the analysis. The red (PwP) and blue (CNT) color scales indicate daily typing activity measured as the number of valid typing windows provided by each subject during the analysis period. We defined a valid window as a sequence of at least 30 keystrokes within 90 s. Panel B illustrates the variability in the amount of typing data with an example from a single PwP subject.
Figure 2
The neuroQWERTY platform. This platform was designed to allow for automatic data retrieval of typing data collected at home and remote management by a study coordinator. Operationally, an account in the neuroQWERTY platform was created for each participant in the study. The data-collection software was downloaded and installed in their users’ personal laptop to enable remote data collection. The data, linked to each user account, was encrypted and automatically sent to a remote server through their home Internet connection. The neuroQWERTY platform also implemented an administrator module to provide the study coordinators with an interface to control and visualize participants’ typing activity.
Figure 3
Algorithm pipeline. The figure represents the pipeline to generate a single neuroQWERTY index (nQi) from a stream of typing data. (1) The typing signal is defined as the time series of hold times corresponding to each keystroke within a typing routine. This signal is split by nonoverlapping 90-s windows that the algorithm will evaluate as independent typing units. (2) Only windows with at least 30 keystrokes within the 90-s interval are analyzed. (3) The neuroQWERTY algorithm, previously trained on a separate in-clinic dataset, computes a single numerical score from each independent window. (4) The final nQi is computed as the average of the window-level scores.
Figure 4
Example of the application of the neuroQWERTY algorithm in an in-clinic typing test.
Figure 5
Example of the application of the neuroQWERTY algorithm in the at-home setting. The neuroQWERTY algorithm described in Figure 3 can be used indistinctly to evaluate controlled or natural typing data. This figure represents the at-home typing data and corresponding scores for the same subject shown in Figure 4 (note different time scales used in Figure 4 and Figure 5). Although the uncontrolled activity appears in unpredictable bursts that introduce a high degree of sparsity, our window-based approach allows to analyze the at-home data using the same method applied for the quasi-continuous in-clinic data.
Figure 6
Comparison of raw typing metrics between in-clinic and at-home typing settings. The figure shows the correlation of the raw typing metrics, hold time (HT; time between pressing and releasing a key), and flight time (FT; delay between two consecutive key presses), between in-clinic and at-home settings. Each point represents the metric coordinates (in-clinic, at-home) for each of the 52 participants included in the analysis. Both HT and FT values are very similar independently of the typing scenario, as shown by the correlation coefficient values. These results suggest that the in-clinic task does not alter the way subjects type in comparison with their natural typing at-home, which supports our hypothesis that the neuroQWERTY algorithm, built in an in-clinic setting, could be applied to evaluate motor impairment using the typing data from an uncontrolled at-home setting.
Figure 7
Comparison of neuroQWERTY index (nQi) between in-clinic and at-home typing settings. We evaluated the influence of the typing setting in the nQi scores by applying a similar analysis as described in Figure 6 for the raw typing metrics. Panel A shows the correlation of the nQi scores computed in-clinic and at-home. Panel B includes the results of the Bland-Altman analysis to evaluate the agreement of our method in the two typing scenarios. The black line shows the mean difference (d) and the top and bottom dashed lines show the limits of agreement (LoA, d±1.96×SDd).
Figure 8
Comparison of neuroQWERTY index (nQi) performance between in-clinic and at-home typing settings. Panel A scatterplot illustrates the in-clinic and at-home nQi scores in a patient level. The two black lines represent the classification thresholds computed in-clinic (nQi=0.0473) and at-home (nQi=0.0667). These thresholds were estimated for closest-to-(0,1) cutoff points that maximize sensitivity/specificity pairs. Panel B presents the comparison of the receiver operating characteristic (ROC) curves showing the classification rate for the in-clinic and at-home nQi. The plotted curves are the average result of the bootstrapped ROC analysis and the shadowed areas represent the corresponding CIs [5th-95th]. The statistical significance of the Mann-Whitney U test is estimated to reject the null hypothesis that the two groups, PwP and CNT, come from the same population. It is noted as: P<.001(***), P<.01(**), and P<.05(*).
References
- de Lau M, Breteler M. Epidemiology of Parkinson's disease. Lancet Neurol. 2006;5(6):525–35. -PubMed
- Gelb DJ, Oliver E, Gilman S. Diagnostic criteria for Parkinson disease. Arch Neurol. 1999 Jan;56(1):33–9. -PubMed
- Ramaker C, Marinus J, Stiggelbout A, van HB. Systematic evaluation of rating scales for impairment and disability in Parkinson?s disease. Mov Disord. 2002;17(5):867–76. -PubMed
- Whone A, Watts R, Stoessl A, Davis M, Reske S, Nahmias C, Lang AE, Rascol O, Ribeiro MJ, Remy P, Poewe WH, Hauser RA, Brooks DJ, REAL-PET Study Group Slower progression of Parkinson's disease with ropinirole versus levodopa: the REAL-PET study. Ann Neurol. 2003 Jul;54(1):93–101. doi: 10.1002/ana.10609. -DOI -PubMed
Publication types
MeSH terms
LinkOut - more resources
Full Text Sources
Other Literature Sources
Medical