Facebook tracking pixel
Section Menu


Our goal has been to establish a remote extended monitoring and mobile health system for risk-related stroke measures to proactively provide patients, caregivers, and health professionals with previously unavailable real-time data at the body structure, activity, and participation levels, whereby patient compliance and progress can be monitored and rehabilitation and/or medical intervention may be triggered to support stroke patients’ optimal long-term recovery.


About mStroke

mStroke is a real-time quantitative assessment of stroke rehabilitation using wireless sensors. mStroke system will evaluate recovery of post-stroke patients after they leave the hospital, and will provide trustworthy customized activity analysis and statistical interpretation to support health care providers in delivering improved health services beyond usual stroke care. We anticipate that this system will have a significant impact on stroke rehabilitation (intervention and research) and patients’ long-term recovery. mStroke monitors and evaluates motor control, fall risk, and gait speed of patients post stroke using wearable Bluetooth Low-Energy (BLE) devices.

figures showing the concept of the mStroke application and mobile interface of the app

* Joint project with Dr. Li Yang (Computer Science) and Dr. Nancy Fell (Physical Therapy)


Functional Reach Test (FRT)

The FRT function in mStroke has been tested independently on two groups of healthy adult subjects, totaling 40 people in all. In the first group, only one NODE (positioned on the chest) is used to estimate trunk flexion and torso twist angles. In terms of Mean Absolute Error (MAE), the performance of the FRT function with consideration of trunk flexion and torso twist is 2.93cm in comparison to the clinical benchmark. In the second group, two NODEs are utilized to estimate trunk flexion, torso twist, and thigh movement angles. The performance using two NODEs can be improved by 17.6% compared with the performance using one NODE.

  Austin Harris performing a functional reach test

Performance comparison between clinical benchmark and mStroke 


NIHSS (Motor Arm/Motor Leg)

The National Institute of Health Stroke Scale (NIHSS) is a widely used tool for clinical evaluation of post-stroke patients that is designed to be a quick and reliable measurement of post-stroke patient capabilities. It is comprised of 15 items which are used to evaluate the effect of acute cerebral infarction on the levels of consciousness, language, neglect, visual-field loss, extraocular movement, motor strength, ataxia, dysarthia, and sensory loss. mStroke is designed to administer two of the items from of the NIHSS Stroke Scale: Motor Arm and Motor Leg. Both of these clinical measures center around a patient’s movement and thus are perfect candidates for the mStroke system. Results on 60 subjects each doing 4 tests are shown below.


    ma2               MotorLef


Fall Recognition

To initiate fall recognition in mStroke, we have acquired motion data from 14 healthy subjects. Each subject was asked to perform a total of 21 activities randomly chosen from three activities of daily living (reaching up, reaching down, and walking) and four falls with different directions. We demonstrate recognition performances based on different feature selection approaches, supervised learning algorithms, sensor configurations, and the feature numbers. All the recognition accuracies are above 90%.


Fall recognition results table



An important part of physical rehabilitation, especially for stroke patients, is determining how well they walk, known as Gait Analysis. In a typical Gait test, the patient is instructed to walk a set distance at a comfortable speed, using whatever walking implements they find necessary. This walk is timed, and the time is compared to known, normative data for patients of a similar class. Other interesting features may be examined as well: step symmetry (is one leg doing more work than the other?), cadence (how quickly are steps taken?), stride and step length, and so on.


Smart Health

Healthcare is the diagnosis, treatment, and prevention of disease, illness, injury, and other physical and mental impairments in human beings. As the sensor and communication technologies, like wearable technologies (e.g., smart watches, fitness trackers, and sleep monitors), internet of things, and machine learning techniques progress fast these days, more and more relevant technologies are exploited in healthcare to improve performance and outcome. Mainly to transform healthcare from reactive to preventive, proactive, and decision based. Focusing on real-time data from IoT devices and historical data from electrical medical records, we are exploring two different applications in smart health.  One is about human activity recognition using wearable technologies and machine learning algorithms and the other is about the prediction of hospital discharge status for stroke patients in the State of Tennessee.

*Joint project with Dr. Gregory Heath (Health and Human Performance), Dr. Nancy Fell (Physical Therapy), and Dr. Rehan Qayyum  (University of Tennessee College of Medicine at Chattanooga)


Activity Recognition

Human activity recognition identifies different human body movements and estimates the dwell time for each activity which can be widely used for behavior analysis and rehabilitation assessment in the areas of elderly care, health care, assisted living, and athlete training. We perform preliminary investigation of recognizing human activities (including real-time fall detection and its direction) using wearable technologies and machine learning algorithms. From daily activities’ point of view, the most dangerous situation for the post-stroke patients or elderlies is fall. To detect fall and assess the severity of fall as early as possible can enable timely intervention.

In our research, a human subject wears an IMU sensor on the chest to acquire raw accelerometer and gyroscope data for each activity. Five different activities, such as sitting/standing, lying down, going upstairs, going downstairs, walking, and falling (in 4 directions) are considered. After acquiring raw motion data, feature construction is executed. Fifty four statistical features are generated based on different statistics, such as Range, Mean, Absolute Mean, Mean Cross Rate, Zero Cross Rate, Standard Deviation, Covariance, and Mean Trend. Feature selection is performed afterward to find a subset of the original full feature set. Three different filter-based feature selection methods are used here, i.e., kBest feature selection, Relief-F feature selection, and robust feature selection. Finally, activity classification is on duty to recognition different human activities based on the input features built for each activity. 


nodeplace                          humanactivityrecog


Predictive Analytics

We used data analytics techniques to build a predictive model to forecast hospital discharge status of stroke patients in the State of Tennessee. Tennessee Department of Health provided hospital discharge data corresponding to stroke patients. In other words, we want to predict whether a stroke patient should be discharged after hospitalization to home or not home. In this way, we can perform personalized health care recommendation, improve quality of health care, and reduce cost of health care. We consider the following potential predictors, i.e., sex, age, race, ethnicity, stroke type, comorbidities (diabetes, heart disease, hypertension, peripheral arterial disease, chronic kidney disease, hyperlipidemia, stroke, arrhythmia, and depression), bill type, admission type, admission source, and primary and secondary payer classes. Three different machine learning methods, i.e., logistic regression, random forest, and support vector machine, are exploited to build predictive models. The following figures show receiver operating characteristic (ROC) curves of predictive models for derivation set and validation set, respectively.












ROC curve for derivation set.                                                                               ROC curve for validation set.  


Health-Centric Urban Mobility

Due to the advancements in smart and mobile health technologies and systems, an individual’s health condition can be remotely monitored, recorded, and analyzed and personal health status can be reliably predicted. This is the core concept behind health-centric urban mobility. For example, personalized trip planning and transportation modality selection should consider the individual’s detailed health information. It is well known that physical inactivity is a major contributor to the steady rise in rates of diabetes, heart disease, and stroke in the United States. Active transportation, which involves physical activities, should be advocated for health promotion. However, if people utilize active transportation in a polluted environment, their health could be jeopardized. Therefore, our goal is to incorporate environmental awareness into health-centric urban mobility to allow individuals to choose active transportation routes with good air quality and plan trips during the periods of time with good environmental conditions.

This high-impact, cross-disciplinary research concept will blend transportation, health, environment, urban science, computer science, computational science, and data science. Large-scale data from multiple heterogeneous sources will be collected and analyzed to identify innovative, sustainable, and economically viable options for health-centric urban mobility to improve quality of life.
multi-modal diagram based on air quality