With finals concluded this past week, I was able to go home for a few days to see family, but jumped right back down to Boston for one of the best events I have been to as a college student; MIT Hacking Medicine’s Grand Hack. Being the largest healthcare hackathon in the world, I was a bit intimidated and didn’t quite know what to expect. However, from the networking session at the start of the hackathon I realized how much time and care was put into the event. Unlike other hackathons, this event was truly interdisciplinary; there were a wide variety of age groups and professions, from doctors to business professionals to engineers. From the start, I was excited to see what new ideas could arise from working on such a diverse team.
As the event progressed, I was able to work my way on to a team comprised of a Physician from Canada, a Neuroscientist Postdoc from Canada, a recent Yale graduate software engineer, a recent UConn graduate Biomedical Engineer and a current AMD hardware engineer. We began working on an idea that arose from one of my teammates trips to Peru as a medical professional. He spoke of the fact that many of the professionals there had smart phones, while the hospitals didn’t even have basic equipment such as an ECG. Given that heart disease is a top killer in the world, this was astonishing. To see if this was a problem across the world, we also had another teammate who was from Kenya call some of his contacts to see if this was a problem there as well. To our surprise, one of the major hospitals in an area with a population of over 1 million people also did not have an ECG machine, even though every physician and nurse in the hospital had smartphones. Thus, we set out on a mission; prototype an ECG that integrates with a smartphone, and also uses machine learning tech. to provide a diagnosis to fix the ECG scarcity and scarcity of medical professional attention in these developing countries.
We divided up the work, and I worked with another team member to write the backend AI. Throughout the weekend, we were able to pull and process raw, dirty ECG data, and also wrote a machine learning algorithm to train the data to predict either healthy or unhealthy heart rhythms from these cleaned ECG signals. You can see the preliminary code on our Github (https://github.com/jrs33/ECG_HackMIT). Our algorithm consisted of compressing these ECG timeseries using Dynamic Time Warping, and then running K-nearest neighbors on the metric generated from the DTW. We got inspiration for this by reading through literature on how quants in finance classify stock timeseries, and thought it was easily transferable to this healthcare application.
With this backend running, we were then able to combine with our teammates to plan and map out the circuit and costs (only $32 to build!) and also wrote a preliminary business plan. We then pitched, and were incredibly excited to be placed as finalists in the competition. Given the stature of the hackathon, we really think we have something solid going, and are hoping to stay in touch to see if we can build a fully functioning prototype to test out our ideas in either of the two places we have contacts (ie: Peru or Kenya).
Me and my fellow ECG Hack team showing off our unexpected hardware
If you have any questions, advice or interest about getting involved in our project don’t hesitate to contact me. Stay tuned for more updates about how our hopeful prototype goes, and another thanks to the MIT Hacking Medicine team for putting together the best hackathon I have been to!