Unravelling socio-motor biomarkers in schizophrenia

Abstract

We present novel, low-cost and non-invasive potential diagnostic biomarkers of schizophrenia. They are based on the 'mirror-game', a coordination task in which two partners are asked to mimic each other's hand movements. In particular, we use the patient's solo movement, recorded in the absence of a partner, and motion recorded during interaction with an artificial agent, a computer avatar or a humanoid robot. In order to discriminate between the patients and controls, we employ statistical learning techniques, which we apply to nonverbal synchrony and neuromotor features derived from the participants' movement data. The proposed classifier has 93% accuracy and 100% specificity. Our results provide evidence that statistical learning techniques, nonverbal movement coordination and neuromotor characteristics could form the foundation of decision support tools aiding clinicians in cases of diagnostic uncertainty.

Fig. 1

Illustration of the main experimental set-up and of the experimental conditions. a Solo condition of the mirror game, hand movement of the participant along a line is recorded on a computer. b In the leader–follower condition participant is following motion of a computer avatar that is displayed on the screen

Fig. 2

Features based on the solo movement. a Position time trace from a solo recording, blue bars indicate lengths of movement segments in the left and right direction ΔP ₀. b Histogram of the values of ΔP ₀ from panel a. c Time-frequency plot of the wavelet transform of the time series from panel a. Colour indicates power for given frequency at a given time. d GWS computed as mean (over time) of the signal representation in panel (c)

Fig. 3

Features based on the artificial-agent leader—human follower interaction. a Position time traces of a leader (black) and follower (green). b Time-frequency plot of the cross-wavelet coherence. Arrows indicate relative phase (3 o’clock indicates in-phase). Colour indicates degree of "overlap" between wavelet spectra of the two time series. c |ϕ _r (f)|—distribution of absolute phase lag over frequencies, computed as the absolute value of a circular mean (over time) of relative phase angles from the regions encircled by the black lines in panel b. d Estimate of the relative phase ϕ _r (t), computed as circular mean (over frequencies) of the relative phase angles from panel b for frequencies lower than 2 Hz, y-axis in radians. e Histogram of the relative phase ϕ _r (t) from panel d

Fig. 4

Illustrative example of the MDS-space and of the Step 7 of the classification algorithm. a Two first coordinates of the MDS-space computed from the EMDs between distributions of absolute phase lag over frequencies |ϕ _r (f)|. Red dots indicate patients while blue dots indicate controls. Corresponding classification measures are: accuracy = 0.7727, sensitivity = 0.7272, specificity = 0.8181, precision = 0.8000. Panel b shows accuracy and c precision of classification as a function of number of the coordinates. Black crosses show result of classification using the first n dominant coordinates. Green circles show result of the best classification using the set of n coordinates out of 15 dominant coordinates