How we used machine learning to hack Dutch healthcare
The third weekend of april it was time for the second ‘Dutch Hacking Health’ Hackathon.
… It is an event where technology, design, entrepreneurship, healthcare and patients are brought together during a weekend. Hacking Health aims to transform healthcare by connecting teams, which usually rarely collaborate, who search for human-centric solutions to front-line healthcare problems.
More than 500 people participated, divided over seven different locations in the Netherlands. The biggest one was at the ‘Universitair Medisch Centrum’ in Utrecht, which is where we joined. Six excited ‘Igniters’ (people who work at Ignation) jumped in the car, train, or whatever other kind of transportation and headed to the UMC hospital.
The theme of this year’s Hackathon was ‘Patient in the lead’. One of the main criteria for all of the solutions built, was that it should have the potential to empower patients. We started thinking about a possible solution in advance, how can we use our knowledge and expertise to improve healthcare and put the patient in the lead?
We have a partnership with IBM, and use their IBM Cloud platform when- and wherever we can. How can we leverage the AI capabilities of IBM, and create something that would benefit patients? What we came up with was an app that scanned your skin for certain skin diseases. Apps like this already exist, but they all rely on people taking a photo of the skin, which is very time-consuming. By using the IBM visual recognition service, we could scan a person’s skin in real-time. This would significantly decrease the time patients would spend with the app, and thereby also the barrier to use it.
We called the project ‘Check je Vlek’ or ‘Spot the Dot.’ in English
Pitch
With our idea in place, it was time to Pitch it to all the other people in the Hackathon, to see if anyone else was willing to join our team. Two more great people joined us, namely a data scientist and a developer.
Let’s get to work!
Teams are formed, excitement is in the air. Hands are itching to start Designing, Coding, and just making cool stuff in general. But first things first. Further develop and fine-tune the idea.
For this project we decided to go with a simplified Design Thinking approach. For those of you unfamiliar with Design Thinking, I recommend reading this blog.
Discover
In this phase we went looking for questions instead of answers. We have this idea, but does this idea relate to how other people see the problem. So instead of diving into a solution for the problem, we talked to as many people as possible. This can be quite a challenge when you have to do it in a weekend, but luckily the Hackathon’s organisation was very helpful, they put us in touch with the dermatology department so we could ask them some questions.
We also did some pre-work before the Hackathon to see what the competition is doing, but more importantly what they are not doing. So much information came up, next up we had to analyze this information, and synthesize it into something we could actually use.
Define
We tried to cram all of the information we gathered into three questions:
What?
Why?
Who?
To answer these questions we took out the good old Post-Its. Everyone in our team wrote down multiple answer to all of these questions. That’s a lot of sticky notes. So we narrowed it down to two by popular vote. We synthesized this into the following product description:
A tool for measuring skin conditions that is easily trainable by dermatologists, to make it approachable for people in doubt to go see a doctor.
Our second step in the Define phase is coming up with personas. At Ignation we do Persona workshops quite often. So, this was a cakewalk. We created two personas one for the patient, and one for the dermatologist.
In the process of developing these personas we figured that our main point of difference with existing apps, is the fact that, the IBM Visual Recognition model that we create is very easy to train by healthcare professionals. Other apps rely heavily on complicated mathematical models. So, we decided to put the focus on the dermatologist to save time (We called him ‘Danny Derma’).
Develop
Now it’s time to start ideation, and creation! We figured that it would be hard to collect enough data on skin diseases in a weekend. So instead, we decided to show how easy it would be to teach the app to detect changes in the data that is entered. Just to make it a bit simpler we worked with dots of different colors and shapes.
With the dermatologist in the back of our minds we created a very brief customer journey for this dermatologist, with accompanying wireframes. This gives an indication of how a dermatologist would use the app.
Deliver
The delivery phase is the final stretch, it’s time to buckle down and get stuff done. The designers in our team started working on multiple things. First we created a logo, with an accompanying splash screen for the app. Then they also designed all of the screens for the app.
The developers transformed the design into a fully functioning iOS app. This app had a back-end integration with the IBM Visual Recognition Service.
Outcome
All of our hard work paid off. After three days of a lot of coffee, many hours of staring at our computer screen, and brain-racking, the app was finished. Check the demo below.
Just knowing that with our amazing team we created this app in a weekend would have been more than enough gratification.
But we also won an Innovation Award!