Wearable devices offer a promise of immense impact on worldwide global health by offering the potential for non-invasive, constantly vigilant, and low cost monitoring of individual condition and fundamental advances in guiding healthcare. The urgency of this objective for its individual and societal benefits will attract an expanding community of researchers from backgrounds in nearly every field of computing. This paper describes the unprecedented benefits and opportunities for computing research in wearable devices as well as the multidisciplinary challenges that have not been encountered individually or combined together in previous research. This paper is focused on providing guidance to the new community of Healthcare in Computing researchers who will both create a new field and forge transformative solutions for healthcare delivery to a worldwide population.
Clinical decision support systems (CDSS) are widely used to assist with medical decision making. However, CDSS typically require manually curated rules and other data which are difficult to maintain and keep up-to-date. Recent systems leverage advanced deep learning techniques and electronic health records (EHR) to provide a more timely and precise results. Many of these techniques have been developed with a common focus on predicting upcoming medical events. However, while the prediction results from these approaches are promising, their value is limited by their lack of interpretability. To address this challenge, we introduce CarePre, an intelligent clinical decision assistance system. The system extends a state-of-the-art deep learning model to predict upcoming diagnosis events for a focal patient based on his/her historical medical records. The system includes an interactive framework together with intuitive visualizations designed to support diagnosis, treatment outcome analysis, and the interpretation of the analysis results. We demonstrate the effectiveness and usefulness of CarePre system by reporting results from a quantities evaluation of the prediction algorithm and a case study and three interviews with senior physicians.
External ventricular drainage (EVD) is a high-risk medical procedure that involves inserting a catheter inside a patient's skull, through the brain and into a ventricle, to drain cerebrospinal fluid relieving elevated intracranial pressure. Once the catheter has entered the skull its tip cannot be seen. The neurosurgeon has to imagine its location inside the cranium and direct it towards the ventricle using only anatomic landmarks. The EVD catheter is thin and thus hard to track using infra-red depth sensors. Traditional optical tracking using fiducial or other markers inevitably changes the shape or weight of the medical instrument. We present a novel method that uses augmented reality (AR) to depict the catheter for EVD and a new technique to precisely track the catheter's tip inside the skull. Our technique uses a new linear marker detection method that requires minimal changes to the catheter and is well-suited for tracking other thin medical devices that require high-precision tracking.
Timely detection of an individual?s stress level has the potential to improve stress management, thereby reducing the risk of adverse health consequences that may arise due to mismanagement of stress. Recent advances in wearable sensing have resulted in multiple approaches to detect and monitor stress with varying levels of accuracy. The most accurate methods, however, rely on clinical-grade sensors to measure physiological signals; they are often bulky, custom-made, and expensive, hence limiting their adoption by researchers and the general public. In this paper, we explore the viability of commercially available off-the-shelf sensors for stress monitoring. The idea is to be able to use cheap, non-clinical sensors to capture physiological signals, and make inferences about the wearer?s stress level based on that data. We describe a system involving a popular off-the-shelf heart-rate monitor, the Polar H7; we evaluated our system with 26 participants in both a controlled lab setting with three well-validated stress-inducing stimuli and in free-living field conditions. Our analysis shows that using the off-the-shelf sensor alone, we were able to detect stressful events with an F1 score of 0.81 in the lab and 0.62 in the field, on par with clinical-grade sensors.
In the paper, we explore the feasibility of monitoring outpatients using Fitbit Charge HR wristbands and the potential of machine learning models to predict clinical deterioration (readmissions and death) among outpatients discharged from the hospital. We developed and piloted a data collection system in a clinical study which involved 25 heart failure patients recently discharged. The results demonstrated the feasibility of continuously monitoring outpatients using wristbands. We observed high levels of patient compliance in wearing the wristbands regularly and satisfactory yield, latency and reliability of data collection from the wristbands to a cloud-based database. Finally, we explored a set of machine learning models to predict deterioration based on the Fitbit data. Through 5-fold cross validation, K nearest neighbor achieved the highest accuracy of 0.8667 for identifying patients at risk of deterioration using the data collected from the beginning of the monitoring. Machine learning models based on multimodal data (step, sleep and heart rate) significantly outperformed the traditional clinical approach based on LACE index. Moreover, our proposed Weighted Samples One Class SVM model with estimated confidence can reach high accuracy (0.9635) for predicting the deterioration using data collected within a sliding window, which indicates the potential for allowing timely intervention.
Healthcare applications supported by the Internet of Things enable personalized monitoring of a patient in everyday settings. Such applications often consist of battery powered sensors coupled to smart gateways at the edge layer. Smart gateways provide an opportunity for implementing local closed-loop optimization of energy consumption in the sensor layer. To implement efficient optimization methods, context of patients need to be considered to adjust energy to demanded accuracy. We propose two approaches, myopic and MDP to consider energy constraints and risk factor requirements. Vital signs, including heart rate, respiration rate and oxygen saturation, are extracted from a Photoplethysmogram signal and errors of extracted features are compared to a ground truth, modeled as a Gaussian distribution. We control the sensor's energy to minimize the power consumption while meeting a desired level of satisfactory detection performance. Compared to non-adaptive methods, myopic reduces an average of 16.9% in energy consumption with the maximum probability of abnormality mis-detection of 0.17 during 24 hours. We demonstrate that MDP can extend the battery life on average of more than 2x with the same average mis-detection probability comparing to myopic during four weeks. We compare, myopic, MDP and non-adaptive methods for 14 subjects in one month.