The students of MSTI Cohort 2 recently passed the second milestone of their launch projects. The Launch Project is a culmination of months of joint technology development work between a student group and industry partner, focused on a full product or solution lifecycle. The milestones are designed to serve as a status update for their classmates and professors, while also allowing groups to set up their final six weeks of work before their launch project final presentations in early December. Ten teams presented their findings on projects ranging from advances in audio technology to augmented reality mapping in real-time. In this post, learn more about two groups pushing technological changes to improve the lives of people with unique medical challenges: teams Uniband and A-Eye.
Parkinson’s disease affects seven to 10 million people worldwide and is the second most common age-related neurodegenerative condition behind Alzheimer’s disease. It is characterized most commonly by hand tremors, limb stiffness, weakness and slowing of movement, and general imbalances. The root cause is generally unknown and there is no cure. It affects people by changing the dopamine-producing neurons in the brain, and clinical treatment generally involves dopamine therapy to correct the natural imbalance. Traditionally, medical providers rely on self-reported data of tremor duration, frequency, and intensity—combined with self-tracked food intake, exercise, and medicine timing—to adjust dosages. Too much or too little dopamine directly affects the severity and frequency of tremors and rigidity an individual can experience. GIX students Joohyung Kim, Yuanwei Zhang, Gero Bergk, and Yewei Zhang are developing a sensor package and companion app that automatically tracks tremor data for long-term, high-accuracy analysis. We sat down with team member Gero Bergk to learn more.
“There’s just a lack of information around tremors,” he said. “We wanted to open a window into the individual’s lives where they can gather quantifiable data on how their tremors progress over time and what affects them. Finding the right balance of medicine is critical for their quality of life.”
One of the people the team interviewed for their research tracks her husband’s tremors in a large spreadsheet, manually entering the data day after day. “It’s quite a burden, especially if you’re trying to quantify it. For most people with Parkinson’s, it’s just on them to do their own data collection.” This lack of easy, accurate, and long term tracking drove their work.
“We wanted to create a device that would let both individuals and doctors look back at what had a positive effect to reduce the tremors (like different eating schedules, more exercise, or a specific dosage.)”
To solve this, the team created a wearable device (that looks similar to a wristwatch) that contains sensing hardware. The raw data is then transferred to a companion app on the wearer’s phone before being sent to the cloud for analysis.
“Our hardware has an accelerometer. We take the raw information, convert it and clean it through signal processing, and then run custom analysis on it. Tremors usually last between three and eight times per second, so we can isolate when they have a tremor and plot it. The hardware talks to an app we’re building, which handles medication reminders, notes about food and exercise, and anything else they want to log,” said Gero.
Long-term trends are crucial to understanding the impacts that variables are having on the person, so the app automatically compares tremor severity to food timing, medicine, and more.
And what’s the team dynamic like? “I think having our different backgrounds has been really beneficial to our team. No one can be good at everything or have the time for everything. It’s beneficial that everyone has different interests and can focus on the things they’re good at.”
Between milestone one and two, the team had to abandon initial work because of unexpected hardware challenges and feedback from users. Changing direction is not unusual during the innovation process. “We had five different options on the table. I’ve just finished designing the new PCB, and our developers are working on the new app,” said Bergk. “We didn’t want to make a prototype that was just hacked together; it didn’t align with our goal, which was to help these people. We wanted something robust, affordable, and reliable.”
Diabetic retinopathy, glaucoma, and macular degeneration are ailments of the eye that are becoming more prevalent in rural areas of the world, and current screening methods are either too expensive or require skilled operators. The World Health Organization reports that 57 million people living in India will be affected by diabetes by the year 2025; of those, 11-20 million may experience diabetic retinopathy as a result. All of these conditions result in severe loss of quality of life, and can ultimately progress to full vision loss. Early, affordable screening is critical to preventing these conditions from advancing. In partnership with Microsoft’s AI for Good initiative, Team A-Eye set out to create an affordable, easy-to-use fundus camera that can be deployed to rural parts of the world and provide frontline screening. By using infrared light with a simple setup, the team’s camera can photograph the interior (fundus) of the eye to enable technicians to screen for abnormalities.
“Three hundred seventy million people around the world suffer from diabetic retinopathy, glaucoma, or macular degeneration,” said team member Steven Han. “It’s a huge problem. In terms of imaging, it’s challenging to get an image of the inside of the eye. You have to aim your camera at the exact right place, focus the eye perfectly, and get the light exactly where you want it.”
Currently, there are two main options for screening: high quality, expensive equipment that can’t be deployed en masse in rural settings, and low tech, low precision lenses that require a high degree of operator skill to use. The team’s solution is attempting to fill the gap: lower cost hardware that requires less skill to operate but still produces meaningful imagery that can be sent to doctors off-site for analysis. “We imagine this in rural eye camps where you have volunteers who collect images and screen the results,” said Rutuja Jadhav. “The users are not medical professionals, but you don’t need to be a doctor to perform the screening.”
The team has worked with their GIX consortium member, Microsoft, to develop a fundus camera roughly the size of a shoebox. A combination of lenses, light filters, mirrors, and a camera, the hardware pairs with their custom software to image the inside of the eye. By using machine learning and computer vision enabled through Microsoft cloud computing, the team intends to automatically stitch multiple images together to form a complete image.
“The biggest challenge we have left is to get everything to work without dilation. If you use visible light to see inside the eye, your pupil contracts and the camera can’t see anything,” said Han. “We want to use infrared so that doesn’t happen.” Adele Parsons added, “We took a step back and focused on solidifying the optical design. We worked with the principal architect from the Microsoft HoloLens team (Joel Kollin) who gave us direction. We used his software to model light paths.”
The A-Eye team is currently partnering with UW Ophthalmology for clinical testing of their device. This allows them to directly compare the quality of their images to those of a clinical quality fundus camera.
As of now, the camera can photograph individual sections of the eye with dilation. Going forward, the team is focused on reducing glare in images and photographing the fundus without dilation.
“The problem that Microsoft came to us with showed that the target population is people who live in India,” said Jadhav. “One of the leading causes of blindness is diabetic retinopathy; if it could be treated in the early stages, there’s a 95% chance of saving the vision. As an Indian, it was a very rewarding experience to work on something that was directly beneficial to my roots.”