The objective of this work is to put in numbers the degree of symptoms severity based on the social behavior and cognitive functioning of mental patients when conducting a routine conversation with their attending doctor. Examination of patient’s facial expression manifestations can be a key indicator towards the quantization of cognitive impairment in respect to receiving external emotion expressions. Recent advancements in computer vision machine and deep learning techniques allow the evaluation and recognition of temporal emotional status through facial expressions. In this context, the paper studies the application of these techniques for the automated recognition of Positive and Negative Syndrome Scale (PANSS) indicators by means of extracting features from patients’ facial expressions during video teleconferences. The paper discusses the technical details of the implementations of a video classification methodology for the prediction of schizophrenia symptoms’ severity, introduces a novel approach for the interpretation of video classification results and presents initial results where it is demonstrated that the proposed automated techniques can classify to a certain extend specific PANSS indicators.