Dynamic facial expression recognition deficits in schizophrenia: A classification study

Schizophrenia patients have stable and multi-dimensional facial expression recognition deficits, which affect their social function. Current clinical early screening methods for schizophrenia are limited. This study aimed to explore schizophrenia patients' expression recognition deficits and develop a discrimination model via machine learning, offering a new way for early screening and precise treatment.

Fifty-one patients with schizophrenia and sixty-two healthy controls completed a task where a face morphed from neutral to full intensity of six basic emotions over 21 s. Participants identified emotions by pressing a key and labeling the emotion. Clinical symptoms of patients were assessed using the Positive and Negative Syndrome Scale, the Scale for the Assessment of Negative Symptoms, the Scale for the Assessment of Positive Symptoms, and the Brief Psychiatric Rating Scale.

Schizophrenia patients had significantly lower accuracy in recognizing fear, disgust, anger, and happiness, and lower sensitivity to fear, sadness, and happiness than healthy controls. They also had trouble recognizing negative emotions and often confused happiness with them. Symptom severity negatively correlated with the facial expression recognition ability. The screening model achieved 79 % accuracy and an AUC of 0.84, with potential for 0.88 sensitivity and 0.89 specificity when indicators were optimized.

This study confirms that schizophrenia patients exhibit significant deficits in facial expression recognition accuracy, error patterns, and sensitivity, all of which are negatively correlated with the symptom severity. The screening model based on these indicators achieves a high AUC value and sensitivity, showing potential as a convenient auxiliary tool for the early screening of schizophrenia.