Objective identification and analysis of physiological and behavioral signs of schizophrenia

Abstract

Background: A patient’s physical activity is often used by psychiatrists to contribute to the diagnostic process for mental disorders. Typically, it is based mostly on self-reports or observations, and hardly ever upon actigraphy. Other signals related to physiology are rarely used, despite the fact that the autonomic nervous system is often affected by mental disorders.

Aim: This study attempted to fuse physiological and physical activity data and discover features that are predictive for schizophrenia.

Method: Continuous simultaneous heart rate (HR) and physical activity recordings were made on 16 individuals with schizophrenia and 19 healthy controls. Statistical characteristics of the recorded data were analyzed, as well as non-linear rest–activity measures and disorganization measures.

Results: Four most predictive features for schizophrenia were identified, namely, the standard deviation and mode of locomotor activity, dynamics of Multiscale Entropy change over scales of HR signal and the mean HR. A classifier trained on these features provided a cross-validation accuracy of 95.3% (AUC = 0.99) for differentiating between schizophrenia patients and controls, compared to 78.5 and 85.5% accuracy (AUC = 0.85 and AUC = 0.90) using only the HR or locomotor activity features.

Conclusion: Physiological and physical activity signals provide complimentary information for assessment of mental health.

• Activity
• heart rate
• multiscale entropy
• schizophrenia
• transfer entropy

Acknowledgements

The authors would also like to thank Proteus Digital Health for providing the data used in this study. YB is an employee of Proteus Digital Health.

Declaration of interest

This research is funded by a Research Council UK, grant EP/G036861/1 (CDT in Healthcare Innovation), a Wellcome Trust Centre Grant No. 098461/Z/12/Z (Sleep, Circadian Rhythms and Neuroscience Institute), and EPSRC grant EP/K020161/1 (Multiscale markers of circadian rhythm changes for monitoring of mental health). The authors declare no conflicts of interests.