Assessing Consciousness in Patients With Disorders of Consciousness Using a Facial Action Unit Intensity Estimation Method Based on a Hybrid Similar Feature Network

Assessing consciousness in patients with disorders of consciousness (DOCs) is frequent and crucial in clinical examination. However, current mainstream methods based on behavioral scales are time-consuming and prone to high misdiagnosis rates. Facial expressions holding crucial cues related to consciousness may offer a convenient and objective means to evaluate their consciousness bypassing motor impairments. In this article, a consciousness assessment method for DOC patients based on facial expression was designed. Specifically, a hybrid similar feature network (HSFN) that integrates multiple features generated by a face alignment network (FAN) and a heatmap regression network was proposed for action unit (AU) intensity estimation. The HSFN was adopted to detect relevant AUs induced by painful stimuli and quantify facial expression changes using the Prkachin and Solomon pain intensity-revised (PSPI-R) formula. DOC patients were ultimately evaluated based on the PSPI-R score. Launching the test on 24 minimally conscious state (MCS) patients and 26 vegetative state (VS) patients, we validated the significant difference in average PSPI-R scores between MCS and VS, and we suggested a threshold of 3.5 for PSPI-R. Among 26 VS patients, six patients exceeded this threshold, with four of them showing improvement in the coma recovery scale-revised (CRS-R) assessment after three months. The proposed method has the advantages of easy operation, rapidity, and low cost and has broad prospects for assisting in evaluating DOC patients in clinical diagnosis.