Breathe Hackathon


Helping people w/
COPD diagnose

COPD Foundation, Imperial College, MIT Media Lab

Dave King, Hemang Rishi, Jun Kamei, Phil Noonan




The Breathe Hackathon, held over a weekend at Imperial, MIT, and Technion, challenged participants to improve the lives of people with chronic obstructive pulmonary disease or COPD, the fourth deadliest disease in the world.


Understanding COPD

We started with an expert briefing on COPD from doctors, nurses, and scientists. We learned that every 4 minutes 1 person with COPD dies, that the people most at risk are seniors in developed nations and children in developing countries, and that exacerbations are the most common cause of death for people with COPD. 


User Journey Mapping

A journey map helped us understand the main pain point for the patient, self-diagnosis. June, a 60-year-old retired teacher suffers from severe COPD. Now and then June’s chronic cough worsens. Like most people with COPD, June isn’t sure if her worsening cough is an exacerbation until she finds herself in a hospital bed. At least, this time, it wasn’t deadly. June could have taken her rescue pack and recovered without a hospital visit, but if it weren't an exacerbation, the pack would do more harm than good. With COPD, every cough is a matter of life and death.


Visualizing Coughs

Once we identified our pain point, we brainstormed solutions, evolved them using lateral thinking and selected a cough detection concept to prototype. The idea was to distinguish coughs from one another and diagnose exacerbations early on.

The prototype is made up of two main parts—machine learning software that listens passively for coughs using the mic of any phone, and a mobile interface. We learned that computers have a hard time distinguishing cough sounds. To work around it, we took the audio data of coughs and visualized it, then used existing image based recognition to predict exacerbations before they happened.


UX Testing

We also created a mobile UI to set up the software, alert users when an exacerbation is detected, and track other symptoms in place of a health journal. With the UI mockups and some of the machine learning code complete, we were ready to hear what real people with COPD thought of our prototype. We selected 5 participants for testing and learned that while they found the idea promising, there was a gap we hadn't addressed in trust. Participants mentioned that they would require some form of reassurance before believing the diagnosis was accurate and risking their lives. We addressed the issue in subsequent tests by changing our flow and having their physician recommend the app.



We didn't win the hackathon, but we presented the results at the European Respiratory Society International Congress conference in Amsterdam to thousands of respiratory physicians and released the software open source.