Objective identification and analysis of physiological and behavioral signs of schizophrenia

Figures & data

Table 1. Demographic and clinical characteristics of study participants.

	Schizophrenia	Controls
Age (Mean ± SD)^b	45.1 ± 12.3	51.7 ± 8.8
Gender^c	58% male	75% male
Medication	All medicated^a	N/A
Length of disease  (years)	6.4 ± 3.4	N/A
Symptoms	All in relative symptomatic remission	No history of mental disorders

^aSchizophrenia patients were prescribed anti-psychotic medication, including Olanzapine, Risperidone, Aripiprazole, Perphenazine, Fluphenazine, Ziprasidone, Haloperidol and Quetiapine.

^bDifference between age distributions of schizophrenia and control groups is not significant (p = 0.09) using a two-sided Student t-test at a 95% confidence interval.

^cDifference between gender distributions of schizophrenia and control groups is not significant (p = 0.31) using Fisher’s exact test.

Figure 1. Adhesive activity and heart rate monitoring patch used in the study (Proteus Digital Health, Redwood City, CA).

Figure 2. First five values of multiscale entropy of heart rate (left) and locomotor activity (right) signals of schizophrenia patients and normal controls. Solid lines represent fitting a third-degree polynomial in a least squares sense into the first five MSE scales, with coefficients presented in Table 2.

Figure 3. First five values of transfer entropy of heart rate (HR) and activity signals of schizophrenia patients and normal controls. Transfer Entropy from HR to locomotor activity is on the left and from locomotor activity to HR on the right. Solid lines represent the fitting of a third-degree polynomial in a least squares sense into the first five TE values. Resultant coefficients are presented in Table 2.

Figure 4. ROC curves and AUC for evaluated models based on locomotor activity only (ACT features; dashed line), heart rate only (HR features; dot-dashed line) and combined HR and activity features (solid line). The AUC values are significantly different between models at a 1% significance level according to two-sided Wilcoxon rank sum test with the Bonferroni correction for multiple comparisons.

Table 2. Features of HR and locomotor activity of schizophrenia and control subjects (mean ± STD, arbitrary units for activity and bpm for HR) together with mRMR ranking^a.

	Heart rate			Locomotor activity
	Schizophrenia	Controls	Rank	Schizophrenia	Controls	Rank
Mean	82.01 ± 10.46	76.11 ± 6.19	4	0.04 ± 0.02	0.06 ± 0.01	15
Median	79.33 ± 10.83	74.70 ± 7.10	27	0.01 ± 0.00	0.02 ± 0.00	17
Mode	72.27 ± 12.71	69.23 ± 8.19	20	0.01 ± 0.00^c	0.01 ± 0.00^c	2
STD	13.95 ± 2.70	14.11 ± 2.23	21	0.08 ± 0.03	0.13 ± 0.03	1†
IQR	18.50 ± 5.00	19.71 ± 4.25	29	0.02 ± 0.02	0.03 ± 0.01	8
L5	1952.19 ± 249.99	1831.15 ± 162.50	12	1.17 ± 0.51	1.61 ± 0.33	9
M10	4311.61 ± 599.30	4016.32 ± 322.46	7	7.15 ± 2.71	10.05 ± 5.25	13
RA	0.38 ± 0.03	0.37 ± 0.02	18	0.71 ± 0.08	0.70 ± 0.06	33
IS	0.30 ± 0.14	0.36 ± 0.12	23	0.16 ± 0.04	0.18 ± 0.03	16
IV	0.60 ± 0.21	0.70 ± 0.23	19	1.23 ± 0.13	1.23 ± 0.23	28
MSE(1)^b	<0.01	<0.01	5	<0.01	<0.01	31
MSE(2)^b	−0.02 ± 0.12	0.05 ± 0.13	24	−0.02 ± 0.04	−0.01 ± 0.03	35
MSE(3)^b	0.09 ± 0.34	−0.20 ± 0.36	14	0.13 ± 0.13	0.12 ± 0.09	26
MSE(4)^b	1.78 ± 0.38	2.35 ± 0.34	3†	0.30 ± 0.11	0.31 ± 0.09	30
TE(1)^b	<0.01	<0.01	10	<0.01	<0.01	32
TE(2)^b	0.02 ± 0.04	0.00 ± 0.03	22	0.00 ± 0.06	0.01 ± 0.04	36
TE(3)^b	−0.07 ± 0.11	−0.01 ± 0.07	6	0.01 ± 0.17	0.00 ± 0.13	25
TE(4)^b	0.19 ± 0.15	0.11 ± 0.05	11	0.18 ± 0.11	0.17 ± 0.12	34

^aSignificantly different variables (p < 0.05) according to two-sided Wilcoxon rank sum test using the Bonferroni correction for multiple comparisons are marked with bold font and †.

^bMSE(x) and TE(x) denote coefficients of a third-degree polynomial fitted in a least squares sense into MSE and TE values.

^cValues are close to zero, but distributions of schizophrenia subjects and controls barely overlap, providing high feature ranking.