TBME-01699-2018 1 * Jorge E. Caviedes, Sr. Member, IEEE, * Baoxin Li, Sr. Member, IEEE, and * Varun C. Jammula Abstract—Spine conditions and disorders manifested as back pain affect nearly everyone at some point in life. Physical therapy is an evidence-based treatment for non-specific chronic back pain, and monitoring spine posture during unsupervised therapeutic exercise is essential to achieve optimum results. Our research builds upon the progress in wearable sensor technology, spinal posture monitoring, and biomechanical biofeedback methods to address the need for monitoring compliance and correctness and support data collection to enable improved assessment of the effectiveness of physical therapy for spinal conditions. We designed and tested a minimalistic stretch sensor array, including best method to mount the sensors, and conducted proof-of-concept tests to demonstrate the feasibility of an end-to-end system including sensor setup, data capture, and real-time pattern recognition algorithm. Results for exercise correctness detection show encouraging sensitivity and specificity values obtained from multiple tests using a simulator and two sensor mounting options. Values ranged between 70-100% sensitivity and 100% specificity. A companion mobile app performed flawlessly, matching the results of the off-line setup while demonstrating real-time feedback (auditory and visual) after each exercise completion. The work is a major step towards a large study and future clinical trials which would lead to a new support system for treatment of low back pain (LBP) and other spinal conditions. The most significant transformative effect on physical therapy for the at-home component is that our method has the potential to solve the challenge of monitoring compliance and correctness with benefits across the board from individual patients to the general field of physical therapy. Index Terms—biofeedback, soft sensors, spine posture monitoring, stretch sensors, therapeutic exercise, wearable sensors I. INTRODUCTION pinal exercise physiology deals with improving strength, flexibility, and stability through therapeutic and fitness exercises [1]. Monitoring compliance and progress for therapeutic and fitness exercise remains a challenge given that accurate data can only be obtained in strictly supervised or highly instrumented environments. After training a subject on a given exercise routine, a physical therapist or coach can visually verify whether the exercise has been performed correctly. Moreover, when the subject is on his/her own, an automated method to monitor execution, track progress and provide real-time feedback is not readily available. One study * Authors are with Arizona State University, Tempe, AZ, USA (correspondence e-mail: jorge.caviedes@asu.edu). reported that even though each exercise was described to the patients by visually performing it, and a descriptive brochure was given to each participant, the exercises were still performed inaccurately [2]. Wearable flexible sensors are presently used to monitor human activity and human health. It is possible to monitor physiological parameters as well as biomechanical activity using pressure and strain sensors. Wireless sensor networks allow data capture and transmission according to standard protocols. For a review of wearable sensors technology, see [3]. Monitoring spine posture and motion has been tackled using three types of sensors: elongation, bend and inertial sensors. Papi et al. conducted a systematic review of the state of the art of current use of wearables for the assessment of spine kinematics and kinetics [4]. The results show that there is a broad variety of sensor types, number of sensors and placement options. The review also indicates that the small number of publications found on portable assessment of spine motion revealed limited adoption of this technique, which is at an early stage of development and translation. The authors concluded that data logging and processing, systems design and fixation are areas to be improved to fully exploit the wide applicability of wearable technologies. Wearable sensors may offer a way to monitor exercise correctness and provide biofeedback. Such approach has been reported in an early paper by Dworkin et al. [5]. The system consists of two elastic loops, one to measure spine length (loop A) and the other (loop B) to measure and correct for the effect of breathing noise as shown in Fig. 1. Fig. 1. Two-loop wearable system for posture monitoring [5]. Wearable Sensor Array Design for Spine Posture Monitoring During Exercise Incorporating Biofeedback S