Full Program »
74. Using optical flow to quantify movement in response to emotional stimuli among people with schizophrenia: a pilot study
Background: People with schizophrenia demonstrate deficits in nonverbal expressivity that have important implications for social interactions. Previous work in the field has typically relied on clinician ratings to measure symptoms associated with schizophrenia. In recent years, advances in computer vision techniques haves led to an increase in novel objective, automated measures to assess nonverbal behavior. In this pilot study, we explored the relationship between automatically quantified evoked movement, clinician-rated symptoms, and antipsychotic medication use.
Methods: Men with schizophrenia (n=19) viewed brief, emotionally evocative videos while being video recorded. We estimated individual participants’ movement using optical flow, the pattern of apparent motion of an object between consecutive frames, from these recordings using the Farneback method. Outputs of this algorithm are vector fields of estimates of the spatiotemporal changes resulting from motion. As an indicator of the intensity of a participant’s motion, we calculated motion energy of each recording by framewise summation of the length of each vector. We examined correlations between average motion energy and symptoms, as well as correlations between average motion energy and antipsychotic medication use measured in chlorpromazine equivalents (CPZ).
Results: Motion was inversely correlated with negative symptom severity for both positively (r = -0.507, p=0.026) and negatively (r= -0.456, p=0.049) valenced stimuli. Motion was correlated with positive symptom severity for negative (r=0.516, p=0.023) but not for positively valenced stimuli (p=0.115). Lastly, motion was not significantly correlated with CPZ scores for either positive or negative valenced stimuli (p’s>0.05).
Conclusions: Automated methods of measuring movement are potentially valuable tools that may efficiently capture meaningful differences in nonverbal expressivity. However, these preliminary findings in a small sample should be interpreted with caution; further study of these methods in schizophrenia is necessary.