What It Takes to Get a GoodKnight’s Sleep

Christian Aguilar, Madison Caldwell, Matt O’Brien, Hadeel Abulenin, Joe Shin, Preetam Vuppituri

Research Poster

Introduction

Christian

I remember back when I was a senior in high school and when everyone would ask me what I wanted to do in the future, I would always tell them: I just want to help people. I didn’t know how I was going to do it, or what I really wanted to help people out with, I just knew I wanted to make a difference in someone’s life. Fast forward three years later and I’m in a meeting with a sleep medicine specialist at the Ann Arbor Veterans Affairs Healthcare System and they tell me “I’m proud of what you guys are doing. This idea definitely is a little bit out there, but if you can make this thing work, you can help change thousands of peoples lives”. Throughout my time in college, I always had a feeling of uncertainty about what I was doing in life but after hearing those words from a medical professional about something that I had helped contribute towards, I knew then that this is how I would be able to help people.

There was a lot of work, time, people, and passion that brought me and my team to this meeting and it was all thanks to a fairly new club at the University of Michigan called MedLaunch. MedLaunch is a multidisciplinary design team that works with community partners in the local Ann Arbor area. In the past we have worked with people who have diseases or disabilities and we design and build medical devices or assistive technology for them. This year, our team had the opportunity to collaborate with a sleep medicine specialist to help create a form of sleep therapy for patients who suffer from PTSD-induced nightmare disorders. This team, which goes by the name of GoodKnight, has spent countless hours working together to get to this point, but our journey wasn’t always as perfect as it might seem. We all had different experiences overcoming obstacles, some that we predicted, some that seemed to come out of nowhere. These passages come from team members who wished to share their experience of what they did to make sure that people can one day, get a GoodKnight’s sleep.

Challenges and Solutions

Hadeel

When I saw the MedLaunch stand at Festifall with its purpose of helping the community using biotechnology, I knew that it was a club I wanted to spend my four years in. But, I was a freshman who had some experience in coding but nothing outside of a classroom setting, and I couldn’t see how I would have a role in the amazing things that MedLaunch was accomplishing. However, it became clear from the first meeting, that the community in MedLaunch was one that would allow me to grow and learn new skills while also being a part of a team. The majority of the first half of the semester was spent searching for possible or past solutions to problems we were facing, and I was glad to be a part of the research team.

The biggest problem we faced was how to know when someone is having a nightmare. It was hard enough to figure out how to detect sleep in and of itself, because most research papers expect you to have all the background knowledge a PhD in this stuff gives you. So while trying to decipher every other word in the research we were finding, it seemed almost impossible to understand these studies. As it neared midsemester, we realized that nightmares were still a very unknown topic in sleep studies, and there would be no concrete way for us to decide whether someone was in a nightmare or not. So after speaking with a sleep doctor, we decided to instead induce lucid dreaming whenever the user goes into REM sleep.

Our next challenge was to figure out how to decode the EEG data we were getting from the headband, and what thresholds we would decide on for REM sleep. Once again, the lack of PhDs we held seemed to taunt us because EEG data was just as difficult to work with as nightmares. Uncertain data, wavering numbers between people, and potential noise and disturbances in our data all posed problems that we didn’t know how to solve. However, our sleep doctor came to our rescue again, and told us that an EMG would be a lot less harder to work with, given that a chin tone EMG would be at zero when the user is in REM sleep. Currently, we are working on how to put those EMG electrodes on the temple, and use that data to detect REM sleep.

Prototype

Matt

GoodKnight has iterated through two prototypes. In our first prototype, which was developed in late November, we integrated four EEG electrodes into a headband with small speakers over the ears. We had purchased the headband on Amazon, and wired in the EEGs ourselves. Protruding from the headband itself were wires that connected to an external compartment containing the power source, AA batteries, and the OpenBCI Ganglion Board. The ganglion board housed all of the computing power and the OpenBCI interface allowed us to clearly see what signals our first prototype had picked up. In our first design review, the protruding wires were a significant source of alarm for almost all reviewers. We knew this would be the case, and we had plans to compartmentalize the design. At the time, however, we were not sure what would be the best way to do so.

When we arrived back at school after winter break, we began the iteration of our second prototype. We knew we had to address compartmentalization. Additionally, Ritika and I spoke to our Biomedical Circuits and Systems professor, who advised us to use a chin-tone EMG in place of an EEG to identify REM sleep. Our community partner was in accordance with this, and advised that this was a wise way to proceed. Now, you may be wondering why we used an EEG in the first place. At the time, one of our most important design requirements was accurate identification of sleep cycle data, which required the use of an EEG. As we developed the product, we realized identification and reaction to REM was more important than identification of sleep schedules, so we began to recreate the headband. Our second, current prototype covers the eyes, houses all wiring and power components inside the headband, and has EMG electrodes at the temples. We were all looking forward to critiques at the next design review, but unfortunately COVID-19 had other plans. Nonetheless, we are still working to identify how we can make our product even better.

Hardware

Christian

After deciding that GoodKnight would be working with OpenBCI’s Ganglion board and a pair of bluetooth speakers, some members of the hardware team with an electrical engineering background saw that our biggest issue was going to lie in our power source. Our very first prototype used two different batteries and they realized that this would create two different problems: it added more weight to the prototype and it would become an issue to charge two different batteries. Having a background in chemical engineering, I could only really tell you about flow in a pipe or setting up a mole balance around a reactor, so I was surprised to hear that this could be a potential issue. With their experience designing and building circuits, we were able to calculate how much power each circuit board required to function. We disconnected the power sources of the original boards and soldered them to a single Lithium ion battery. I made sure to pay attention to how my other team members were thinking and even gained some experience soldering (something that I haven’t done since I first learned it four years ago).

This moment in the second prototyping stage made me realize a couple things. I first realized that taking a class in physics of electricity and magnetism wasn’t a complete waste of my time. But more importantly, this experience also made me appreciate the vision of MedLaunch. One of the core values that separates MedLaunch from a lot of other design teams is our focus on forming multidisciplinary teams. If it weren’t for my other team members’ diverse background and spotting those problems in our first prototype from the get go, our prototyping timeline would have been detrimentally delayed. Working in a multi-disciplinary team not only prepares us for the real world where we will be working with people from all sorts of fields, but it also allows each of us to learn from others and teach others from our own background and experience.

Backend Software

Joe

For those of you who don’t know anything about EEG analysis, neither do I. Esteemed, well-qualified researchers have conducted sophisticated studies on EEG signals for years, and my study of humans ended with AP Biology. Thankfully, we found an open-source brain-computer interface platform called OpenBCI to help us out.

OpenBCI can do all the heavy processing for us, including calculating the relative strengths of each type of brain wave; what we had to do was edit the source code so that we could store EEG data in real-time to Firebase. After a few hours of experimenting, we were able to isolate the module that got the information that we needed and we were able to add our own code to print out which type of brain wave was most present. However, we realized we couldn’t send this data to Firebase by editing the source code. Even if we could, users would need to have a computer running next to them while they slept to run the edited OpenBCI code.

Along with the change to EMG from EEG, adding the Raspberry Pi to our design allowed the Ganglion board to send EMG data to the Pi so that it could connect and send data to Firebase. We have written a short Python script that receives raw EMG data and stores it in Firebase. Currently, we are working on processing the raw EMG data to extract relevant information and storing it in Firebase.

Creating the App

Madison

Learning in a classroom is very different from solving a problem for a real person. EECS projects are make-believe: professors create them with just enough information to create a challenging, yet solvable, problem. If I hit a roadblock, I can hop onto Piazza and have my question answered within 10 minutes. However, real problems are unpredictable and dynamic. I learned this very quickly while I was working to develop an app for GoodKnight. This app had components that I knew how to create and components that I had never seen before. It required countless hours on Youtube, learning how to cause half of the page to change when a button is pressed or how to let a user upload a photo. All of these things are challenging, but they’re the reality of app development. MedLaunch has allowed me to learn by doing: I use my resources to learn the new skills that I need to make my app. While it can be extremely frustrating at times, at the end of the day I’m proud of myself for the app that I helped to develop. No hints, no Piazza, just Youtube, Stack Overflow, and me.

Learning App Design

Preetam

When I joined MedLaunch, I didn’t know what to expect. The club itself was doing amazing things, working to combine engineering with medical applications, but I had no idea how I, a freshmen with limited coding and engineering experience, could actually contribute. As we were assigned teams and GoodKnight started brainstorming ideas, what struck me immediately was how inclusive and open to ideas everyone on the team was. No one was judged for their crazy ideas or wacky questions, and it became clear that we were all on this journey together — there was a place for everyone on the team and room for us all to contribute. Even though I had limited software knowledge, I was able to contribute from the research aspect and help design presentations and powerpoints. From looking into what electrodes would be best or deciding on what signals we were looking for, I was surprised at how much info from my classes I could actually apply to a real-world project like this. And as the project developed and our idea became more specific, I even got the chance to learn new skills like app-design. Despite exams, homework, and having to deal with their own share of the project, people were willing to take time out of their day to sit down and teach me things like Android Studio. Despite limited Java knowledge from back in high school, I’m currently working on learning enough to be able to make a version of the app for Android devices as well.

Michigan Business Challenge

Matt

I consider myself a good public speaker, but this was the very first time my skills had been put to the test. Christian, Sujai, and myself spent hours practicing and perfecting a business pitch for GoodKnight. This entailed expounding details about the functionality of GoodKnight, the planned business model, related market research, and a timeline detailing the future of GoodKnight. The Challenge began back in November, right around our first prototype iteration. We began in a pool of 100 teams, and we competed in the IP track of the challenge. We had very successful preliminary and first round presentations, and we were one of eight teams going into the semi-final round. Unfortunately, we were eliminated at that stage. It’s important to note that Christian and I were not giving presentations to our peers or University faculty; Christian and I were talking to real, established investors! The entire experience was eye opening to me. It allowed me to not only believe more in my own ability to concisely describe my goals, but to believe more in the future of GoodKnight.

Market Research

Matt

Prior to the Michigan Business Challenge, Market Research was not something myself or other members of the team focused on. However, as Chrisitan and I gained interest in MBC, we realized the importance of strategizing how GoodKnight would be able to scale this device as a marketable product. I conducted some preliminary identification of who we would be targeting with GoodKnight, called the target market. Based on sleep studies and nationwide survey data, I estimated that there are about 81 million Americans that struggle to sleep at night. Out of that population, I estimated that 7.4 million Americans suffer from PTSD related nightmare disorder, and 1.3–2.3 million of those Americans are Veterans with PTSD. These previous two populations comprised our target market: the people on which we aim to have the greatest effect through the creation of GoodKnight. In my eyes, Market Research is much more than simply evaluating who could be affected by a product; it’s thorough research into competing products that are looking to solve the same problem. I researched the sleep aid market, leading products, and its segmentation. Medication is fairly dominant, but there are technological products that have not gained traction in the market. Whether that is due to a lack of efficacy or lack of consumer trust is of utmost importance to us, as these products would be our greatest competition. One of my tasks going forward is to keep an eagle eye on this sector of the market to identify threats to GoodKnight’s success.

Future Steps

Madison

Our prototype is definitely a work in progress. I’m really proud of the progress that GoodKnight made this year, but we have a lot to do going into next year! I’m a member of the app development team, so naturally I am most excited about the user interface updates that we want to implement. After I finish the iOS app, I’ll switch to Android Studio and begin to create an Android app. We also want to improve the accuracy of our product by adding a second biometric, heart rate. To do so, we will eventually utilize the Apple Watch’s heart rate sensors. I’m looking forward to this upgrade because it will give me the opportunity to develop an Apple Watch interface, which is completely new to me!

The software team is currently brainstorming ways to enhance the software in order to improve user experience. Besides adding heart rate to the algorithm, they are also looking into implementing machine learning techniques into the program. This will make GoodKnight a personalized machine for every user; one that can learn someone’s sleep patterns to make it an extremely precise predictor of nightmares. Although these improvements are intensive, I am confident that we have created a strong foundation to build upon. Now that we have the basic structure of GoodKnight developed, it’s time to implement some really cool stuff.

Conclusion

Madison

Our project has challenged every single person in GoodKnight to push their skills beyond what they thought they were capable of. We all tried new things, learned from each other, and pushed our prototype forward. Although we have faced a number of obstacles (including a global pandemic), our willingness to continue developing our product has prevailed, and we have made something new and exciting. This process has taught me so much about what it means to be on a design team and to contribute to a creation. GoodKnight has helped me feel like I have the ability to think creatively, produce something from scratch, and help people. So thank you, to each and every member of GoodKnight, for being the most amazing, inclusive, and inspiring team. Our product might not be ready for market right now, but I know that someday, real life people will be able to get a GoodKnight’s sleep because of this team, and I’m excited to keep working towards this goal! 2020–2021 is going to be a very exciting year for us.

Sponsors