The feasibility of automated detection of the aligned trunk in sitting directly from raw video using a depth camera

Abstract

Introduction Maintenance of an upright head and trunk ('trunk') posture is mandatory for effective performance of everyday activities. Poor trunk control is commonly seen in children with neuromotor or neuromuscular disorders, (e.g. cerebral palsy or muscular dystrophy), and in adults with neurological conditions (e.g. stroke): this leads to functional limitations, such as compromised ability to sit independently [1-3]. Detailed and reliable evaluation of trunk control status is essential when planning therapeutic interventions for these patients [2-5]. The Segmental Assessment of Trunk Control (SATCo) [4] evaluates sitting control at seven separate trunk segments, judging each segment's position in space relative to a defined, neutral aligned posture. SATCo is a validated test that is in regular clinical and research use but it remains a subjective assessment. Generating objective, automated tools that are feasible for use in a clinical environment, will positively impact the planning and delivery of therapeutic interventions for both children and adults. Research Question This study tests the feasibility of deep convolutional neural network (DCNN) analysis of raw high definition and depth (HD+D) video to identify those video frames containing the aligned, reference trunk posture.