COVID-19 took the world by surprise. As Bill Gates warned us through his Ted Talk in 2015, we are not ready for an epidemic. Furthermore, if there were millions of deaths in the coming decades, they wouldn't be from missiles, but rather from microbes. However, an effective response system can be developed to combat pandemics like COVID-19 and reduce deaths and cases throughout the globe. In our community, inaccurate information diffuses rapidly which leads to illogical action taken as COVID-19 spreads. Also, certain populations, such as my grandfather who recently suffered a stroke, are more vulnerable to the virus than others, thus creating a need for these types of cases to be addressed. There are millions of people who fall within the category of being "vulnerable" to the virus. For now, we need to identify these populations and provide them with the resources and education they need to mitigate spreading and deaths. I want to fix these problems by developing a promising method to identify vulnerable populations and enable them to get the tools they need. I also want to implement a personalized system that advises the user about precautions that should be taken based on the severity of the disease in their region.
What it does
ProActive intends to develop a cost-effective yet reliable form of vital extraction through optical photoplethysmographic (PPG) techniques based on the variances of the color in the blood in order to identify vulnerable populations to COVID-19. Traditional PPG techniques have only discussed obtaining heart rate or oxygen saturation by placing a finger on a single camera. The proposed solution involves the unique usage of the front and rear cameras to obtain a detailed PPG waveform for also estimating heart rate variability, blood pressure, QT interval, and other vital measurements. This novel prototype was able to accurately determine heart rate, with accuracy over 97%, and blood pressure, with accuracy over 93%. What's incredibly unique about ProActive is that it utilizes AI and Machine Learning along with a simple smartphone camera to detect vulnerable populations. Essentially, the user would scan his or her finger with the camera which, through image segmentation is able to decipher informative vitals. Based on the vital sign diagnosis, the smartphone app is able to determine if the person being tested has a higher chance of being vulnerable to COVID-19, thus serving as an effective way to identify vulnerable populations.
I imagine ProActive to be a smartphone app that gathers and compiles accurate information about COVID-19 through organizations such as CDC and WHO. When a user downloads the app, he or she would be able to see specific information about the virus relating to their specific region. Based on the regional data and news, the user would be advised of precautions that should be taken such as hygiene practices, social distancing, and other preventative measures. Identified vulnerable populations would be connected with point-of-care tools for speedy detection of COVID-19. For example, Abott ID Now is currently the fastest available molecular point of care test that can be used to detect the coronavirus. Through this method, cases can be discovered faster and preventative measures can be taken to slow the spread of the virus.
How I built it
The software was developed for a tablet or smartphone running the Android operating system. The program was tested compared to an FDA approved Juning digital wrist blood pressure monitor. The Software was tested remotely on 6 subjects and when compared to the real-time blood pressure monitor, had an overall accuracy rate of 93% for blood pressure measurements and 97% for heart rate measurements. Based on these 2 vitals, several other vitals where extracted which will were used to identify the vulnerable populations. The algorithms were programmed in Eclipse using image segmentation and Machine learning image processing and then were implemented into Android studio to create the app.
Challenges I ran into
I was able to successfully implement the algorithm for extracting blood pressure and heart rate via the smartphone camera. However, I wasn't able to finish networking vulnerable populations to resources which is a future endeavor of this project. Furthermore, once vulnerable populations have been identified, I want to connect them with the resources necessary to mitigate the risk of COVID-19 for themselves and their families.
Another challenge that I ran into involved testing of the project. Furthermore, since I don't have any android products and I'm in quarantine, I wasn't able to test the app myself. However, I contacted 6 people with Android devices who agreed to test my system and send me back data so I could tune the parameters of my app.
Accomplishments that I'm proud of
I'm proud of successfully implementing an app that can scan for Heart rate and blood pressure with a high, clinically acceptable accuracy. These vitals can be used to decipher many other vitals as well as identify vulnerable populations that can be connected to resources, hospitals and etc. These vitals can also be potentially used to detect patients who have COVID-19 but the goal, for now, is just to identify vulnerable populations who have heart problems and irregular vitals.
What I learned
In this global fight to combat COVID-19, it's important that we take into consideration how we can optimize resource allocation to prevent more deaths and cases. Based on this, early action needs to be taken to potentially prevent the spread and also mark individuals more vulnerable to the virus. For these vulnerable populations, medical care packages can possibly be sent out.
What's next for ProActive
Future research involves the improvement of systolic and diastolic blood pressure based on the average blood pressure calculations, investigation on aging and arterial stiffness indices and the extrapolation of certain PQRST characteristics of the ECG from the PPG signal for further refining the diagnostic functions of the current prototype and increases the alignment to protocols followed in clinical settings based on the analysis of ECGs. It is also important to determine the impact on changing certain algorithms to run faster while still obtaining accurate results, and finding an ideal blood pressure using either calculated data or a calibration method for centering the systolic and diastolic values from calculated mean blood pressure. All of these metrics will be used to more effectively identify vulnerable populations.
Furthermore, I also want to add a section to the app where accurate COVID-19 is analyzed through a Machine Learning Neural Network to personally advise users of safety practices relating to hygiene(washing hands, etc) and behavior(leaving the house).