Depression is a major health issue affecting over 21 million American adults that often goes untreated, and even when undergoing treatment it is hard to monitor the effectiveness of the treatment. To address these issues, we have created a real-time depression monitoring system for the home. This system runs 24/7 and can potentially detect the early signs of a depression episode, as well track progress managing a depressive illness. A cohesive set of integrated wireless sensors, a touch screen station, mobile device, and associated software deliver the above capabilities. The data collected are multi-modal, spanning a number of different behavioral domains including sleep, weight, activities of daily living, and speech prosody. The reports generated by this aggregated data across multiple behavioral domains are aimed to provide caregivers with more accurate and thorough information about the client's current functioning, thus helping in their diagnostic assessment and therapeutic treatment planning as well for patients in the management and tracking of their symptoms. We present data of a case study showing the value of the system, deployed over a period of two weeks in a home during a depressive episode. Larger scale studies are planned for the future.

https://www.cs.virginia.edu/~stankovic/psfiles/robempath.pdf

Empath: a continuous remote emotional health monitoring system for depressive illness

Traditionally, automatic sleep stage classification is quite a challenging task because of the difficulty in translating open-textured standards to mathematical models and the limitations of handcrafted features. In this paper, a new system for automatic sleep stage classification is presented. Compared with existing sleep stage methods, our method can capture the sleep information hidden inside electroencephalography (EEG) signals and automatically extract features from raw data. To translate open sleep stage standards into machine rules recognized by computers, a new model named fast discriminative complex-valued convolutional neural network (FDCCNN) is proposed to extract features from raw EEG data and classify sleep stages. The new model combines complex-valued backpropagation and the Fisher criterion. It can learn discriminative features and overcome the negative effect of imbalance dataset. More importantly, the orthogonal decision boundaries for the real and imaginary parts of a complex-valued convolutional neuron are proven. A speed-up algorithm is proposed to reduce computational workload and yield improvements of over an order of magnitude compared to the normal convolution algorithm. The classification performances of handcrafted features and different convolutional neural networks are compared with that of the FDCCNN. The total accuracy and kappa coefficient of the proposed method are 92% and 0.84, respectively. Experiment results demonstrated that the performance of our system is comparable to those of human experts.

A New Method for Automatic Sleep Stage Classification

Multimodal data-driven approach has emerged as an important driving force for smart healthcare systems with applications ranging from disease analysis to triage, diagnosis and treatment. Smart healthcare system necessitates new demands for data management and decision-making, which has inspired the rapid development of medical services using artificial intelligence and new transformations in the healthcare industry. In this paper, we provide a comprehensive survey of existing techniques which include not only state-of-the-art methods but also the most recent trends in the field. In particular, this review focuses on the types of decision-making processes used in smart healthcare systems. Firstly, approaches that utilize multimodal association mining with fine-grained data semantics in smart healthcare systems are introduced. We review the smart healthcare-oriented semantic perception, semantic alignment, entity association mining, and discuss the pros and cons of these approaches. Secondly, we discuss approaches for multimodal data fusion and cross-border association that have been employed in developing smart healthcare systems. Finally, we focus specifically on the use of the panoramic decision framework, interactive decision making, and intelligent decision support systems. We introduce how smart healthcare systems can be applied to and benefit a wide variety of fields, including knowledge discovery and privacy protection.

/pdf/a-survey-on-multimodal-data-driven-smart-healthcare-systems-n1f6ngn55z.pdf

A survey on multimodal data-driven smart healthcare systems: approaches and applications

Unobtrusive sleep state measurements in preterm infants - A review.

Current solutions of sleep apnea diagnosis require the patient to undergo overnight studies at a specialized sleep laboratory. Due to such inconvenience and high cost, millions of sleep apnea patients remain undiagnosed and thus untreated. Based on a micro-electro-mechanical systems (MEMS) sensor and an effective apnea detection algorithm, we propose a low-cost single-channel apnea screening solution applicable in the comfort of patients' homes. A prototype device was designed and assembled including a MEMS sensor for measuring the patient's nasal air flows, and a time-domain signal processing IC for apnea detection and autonomous scoring. The IC chip was fabricated in standard 0.5- $\mu{\rm m}$ CMOS technology. The proposed device was tested for both respiratory rhythm detection and sleep apnea screening under clinical environment. Apnea-hypopnea indices (AHI) were scored to indicate severity of sleep apnea conditions. Test results suggest that the proposed device can be a valuable screening solution for the broader public with undiagnosed apnea conditions.

A Home Sleep Apnea Screening Device With Time-Domain Signal Processing and Autonomous Scoring Capability

Sleep monitoring is an important issue and has drawn considerable attention in medicine and healthcare. Given that traditional approaches, such as polysomnography, are usually costly, and often require subjects to stay overnight at clinics, there has been a need for a low-cost system suitable for long-term sleep monitoring. In this paper, we propose a system using low-cost multimodality sensors such as video, passive infrared, and heart-rate sensors for sleep monitoring. We apply machine learning methods to automatically infer a person's sleep state, especially differentiating sleep and wake states. This is useful information for inferring sleep latency, efficiency, and duration that are important for long-term monitoring of sleep quality in healthy individuals and in those with a sleep-related disorder diagnosis. Our experiments show that the proposed approach offers reasonable performance compared to an existing standard approach (i.e., actigraphy), and that multimodality data fusion can improve the robustness and accuracy of sleep state detection.

C.A. Landis

Papers

Multimodality Sensor System for Long-Term Sleep Quality Monitoring