A tablet-based quantitative assessment of manual dexterity for detection of early psychosis

Abstract

Background: We performed a pilot study on whether tablet-based measures of manual dexterity can provide behavioral markers for detection of first-episode psychosis (FEP), and whether cortical excitability/inhibition was altered in FEP.

Methods: Behavioral and neurophysiological testing was undertaken in persons diagnosed with FEP (N = 20), schizophrenia (SCZ, N = 20), autism spectrum disorder (ASD, N = 20), and in healthy control subjects (N = 20). Five tablet tasks assessed different motor and cognitive functions: Finger Recognition for effector (finger) selection and mental rotation, Rhythm Tapping for temporal control, Sequence Tapping for control/memorization of motor sequences, Multi Finger Tapping for finger individuation, and Line Tracking for visuomotor control. Discrimination of FEP (from other groups) based on tablet-based measures was compared to discrimination through clinical neurological soft signs (NSS). Cortical excitability/inhibition, and cerebellar brain inhibition were assessed with transcranial magnetic stimulation.

Results: Compared to controls, FEP patients showed slower reaction times and higher errors in Finger Recognition, and more variability in Rhythm Tapping. Variability in Rhythm Tapping showed highest specificity for the identification of FEP patients compared to all other groups (FEP vs. ASD/SCZ/Controls; 75% sensitivity, 90% specificity, AUC = 0.83) compared to clinical NSS (95% sensitivity, 22% specificity, AUC = 0.49). Random Forest analysis confirmed FEP discrimination vs. other groups based on dexterity variables (100% sensitivity, 85% specificity, balanced accuracy = 92%). The FEP group had reduced short-latency intra-cortical inhibition (but similar excitability) compared to controls, SCZ, and ASD. Cerebellar inhibition showed a non-significant tendency to be weaker in FEP.

Conclusion: FEP patients show a distinctive pattern of dexterity impairments and weaker cortical inhibition. Easy-to-use tablet-based measures of manual dexterity capture neurological deficits in FEP and are promising markers for detection of FEP in clinical practice.

Keywords: Transcranial Magnetic Stimulation (TMS); autism spectrum disorder (ASD); behavioral marker; first-episode psychosis (FEP); manual dexterity; schizophrenia; tablet-based assessment.

Figure 1

Group performance in Tablet task. (A) Finger Recognition task: mean reaction time in for each condition: mirror (filled bar), inverse (hatched), and perpendicular (light filled) for each group of subjects: Control, First-Episode Psychosis (FEP), Schizophrenia (SCZ), and Autism Spectrum Disorders (ASD). Error bars represent 95% CI. Horizontal brackets: significant differences with * = p < 0.05; ** = p < 0.01; *** = p < 0.001. Post-hoc tests revealed that FEP patients performed significantly slower than control and SCZ subjects, and that SCZ and ASD patients also performed significantly slower than control subjects. (B) Rhythm task: mean Intra Subject Variability (ISV) at 3 Hz in the Feedback (dark filled) and NoFeedback (hatched) condition, for each group. Post-hoc tests showed that FEP patients exhibited a larger variability (tapping irregularity) compared to control subjects in the NoFeedback condition and compared to all other groups in the Feedback condition. (C) Line tracking task: mean task duration in Single-task (dark filled) and Dual-task (hatched) conditions, for each group. Post-hoc tests revealed that SCZ patients performed significantly slower than control and ASD subjects, and that FEP also performed slower than control subjects, in both Single and Dual-task conditions. No significant differences were found between the two conditions. (D) Mean Total NSS score for each group. Post-hoc tests revealed that FEP, SCZ, and ASD patients had a significantly higher NSS Total score compared to control subjects. Total score (Kruskal-Wallis, H(3) = 25.90, p < 0.001) (p < 0.001 for SCZ and ASD) (p = 0.047 for FEP). But no significant differences between the three psychiatric groups were found [H(2) = 4.403, p = 0.111].

Figure 2

Radar plots for tablet task performance for each group Z-scores (clock-wise, starting from 12 h) for key variables differentiating group performances: Finger Recognition (FR) task: reaction time, correct trial, and coactivations in inverse condition; Rhythm tapping (RhT) task: inter-tap-interval Feedback, inter-tap-interval Feedback-NoFeedback, and Intra Subject Variability 3 Hz Feedback; Multi Finger Tapping (MFT) task: reaction time, correct trial, and coactivations for double taps; Line Tracking (LT) task: duration, number of velocity peaks, and number of micromovements in single-task condition. (A) Control group: performance data were normalized to 1 (and respective performance of the patient groups expressed as z-scores). Patient groups do not show generally decreased performance (i.e., uniformly increased radius of the plot), but show deficits in particular tasks and variables. (B) FEP patients showed higher (worse) scores in Finger Recognition (FR) reaction time and coactivations, and in Rhythm Tapping (RhT) ISV 3 Hz Feedback. (C) SCZ patients showed higher (worse) scores in Rhythm Tapping (RhT) ITI Feedback, ITI Feedback-NoFeedback, and ISV 3 Hz Feedback. (D) ASD patients showed higher (worse) scores in Finger Recognition coactivations and Multi-Finger Tapping correct trial. * show significant differences from control: * = p < 0.05; ** = p < 0.01; *** = p < 0.001.

Figure 3

ROC curves for tablet performance and NSS score. (A,B) One-vs-One ROC curves for First-Episode Psychosis (FEP) patients vs. control subjects. (A) Rhythm Tapping performance (Intra Subject Variability at 3 Hz Feedback): sensitivity = 75%, specificity = 88.5%, Area Under the Curve = 0.812. (B) Clinical NSS Total score: sensitivity = 90%, specificity = 60%, Area Under the Curve = 0.797. (C,D) One-vs-All ROC curves for FEP patients vs. Control subjects and patients with Schizophrenia (SCZ) or with Autism Spectrum Disorders (ASD). (C) Rhythm Tapping performance (Intra Subject Variability at 3 Hz Feedback): sensitivity = 75%, specificity = 89.5%, Area Under the Curve = 0.815. (D) Clinical NSS Total score: sensitivity = 95%, specificity = 22%, Area Under the Curve = 0.492. These ROC curve results are statistically significant (p < 0.01 or p < 0.001), except for NSS total (p = 0.918) in the One vs. All approach [and for FR coactivations (p = 0.093) not shown].

Figure 4

CART decision tree. Classification of the four groups using CART decision tree algorithm: Control subjects, patients with First-Episode Psychosis (FEP), with Schizophrenia (SCZ), or with Autism Spectrum Disorder (ASD). The 4 successive numbers in the rounded rectangles represent: number of control subjects, of FEP, of SCZ and of ASD patients in a particular group; e.g., if RhT ISV 3 Hz Feedback > = 55 ms ➔ 0 13 0 0 in the FEP group after the first branch corresponds to: 0 (among 29) control subjects (vertical writing), 13 (among 20) FEP, 0 (among 20) SCZ and 0 (among 20) ASD patients were classified as FEP patients (and so on for the following branches). Over the entire tree, false detections thus concerned 4/29 controls, 2/20 FEP, 7/20 SCZ, and 2/20 ASD patients.

Figure 5

Transcranial magnetic stimulation group results. (A) Mean short-latency intracortical inhibition (SICI, %), for the four groups: Control subjects, patients with First-Episode Psychosis (FEP), with Schizophrenia (SCZ), or with Autism Spectrum Disorder (ASD). Error bars and horizontal brackets as in Figure 1. Post-hoc tests revealed that FEP patients had a significantly lower (weaker) SICI compared to Control subjects, to SCZ and ASD patients. (B) Mean Cerebellar Brain Inhibition (CBI, %) of M1 for the four groups. Error bars as in Figure 1. Post-hoc tests did not show any significant group difference in CBI. * show significant differences: * = p < 0.05; ** = p < 0.01; *** = p < 0.001.