milutin stanacevic, PHD

Milutin Stanacevic received the M.S. and Ph.D. degrees in electrical and computer engineering from Johns Hopkins University, Baltimore, MD, USA, in 2001 and 2005, respectively. In 2005, he joined the faculty of the Department of Electrical and Computer Engineering, Stony Brook University, Stony Brook, NY, USA, where he is currently an Associate Professor. His research interests include mixed-signal VLSI circuit design for RF energy harvesting in implantable devices and tag networks, ultra-low power biomedical instrumentation, and acoustic source separation. Dr. Stanacevic is a recipient of the National Science Foundation CAREER award and IEEE Region 1 Technological Innovation Award. He was an Associate Editor of the IEEE Transactions on Biomedical Circuits and Systems and serves on several technical committees of the IEEE Circuits and Systems Society.

ABSTRACT

Wearable Device with Multi-Modal Array Sensing and Deep Probabilistic Learning Methods for Early Noninvasive Identification of Dysphagia

Although the number of proposed wearable devices with intended application in healthcare has been steadily growing, the commercially available devices have mostly been in the form of the smartwatch or wristband and with limited applications especially in the field of personalized diagnostic tools. One of the major obstacles has been validation and reliability of these devices, as well as their long-term performance. We envision a new direction in the design of wearable devices that would address the issues of the reliability and long-term performance of these devices for early detection and diagnostics.

We address the early detection of dysphagia or impaired swallowing. Dysphagia is well-recognized as a cause of aspiration pneumonia, dehydration and malnutrition, prolonged hospitalization and/or hospital readmissions, and death. Early identification and management of dysphagia is critical to reduce the associated medical and personal burdens. The ground truth in assessing the swallowing process is obtained using videofluoroscopic swallowing study (VFSS), or a Modified Barium Swallow Study, which are both imaging techniques that enable identification of specific events during the swallowing process [Han, 2001; Croghan 1994]. However, this is a costly and invasive technique that has to be performed in a hospital. Different pre-screening techniques, typically involving an expert scoring the patient’s swallowing process, are used to determine the necessity of VFSS. However, approximately 40% of patients with dysphagia, do not exhibit any clinical signs of aspiration (e.g. coughing, vocal wetness) during a clinical swallow evaluation. There are no tests that can assist in the diagnosis of aspiration pneumonia in patients without any clinical signs of aspiration.

We propose a multi-modal sensing platform that combines the monitoring of muscle activity and microphone array that captures the acoustic events. The wearable device leverages these high- resolution sensing arrays and deep probabilistic learning methods. High-density electromyography (EMG) recording array can isolate muscle activity which is especially important in the recording of the muscles that participate in swallowing due to size and spacing between these muscles. The microphone array can isolate and localize the specific events in the swallowing process. The fusion of the surface EMG and acoustic signals have previously shown promising results, however only a single sensor for both modalities have been used [Hsu, 2013] which greatly limits the detection and quantification performance of the wearable device. We address the system and circuit implementation challenges that arise from the limited available computational capabilities and communication bandwidth. These resources are severely constrained due the form factor and available energy of the wearable device. These challenges become more pronounced with the use of sensing arrays that call for additional resources both in terms of computation and communication.

To demonstrate the functionality of the proposed device, we present results with a prototype device that has been used for differentiating swallowing while consuming food of different textures. In addition to the use as an early detection device, we also envision use of the device as an in-home monitoring tool that patients can use during meals to assess the risk of aspiration. The monitoring device could significantly reduce the incidence of aspiration pneumonia in the geriatric population.