Inspiration
Due to a lack of educational materials on People of Colour, dermatologists are unable to diagnose PoC adequately. Modern solutions using AI also seem to show systematic biases - while being helpful to diagnose diseases for lighter skinned people, the results for darker skin tones cannot convince. Some projects like Our Vision from Malone Mukwende or AI solutions such as an app from korea spark hope, but don’t tackle this issue in a broad way. As we think a systematic bias like that should be challenged, we decided to start acting on our beliefs and to work on a possible solution.
What it does
SKINai is going to be an app that allows dermatologists to classify and upload images of skin diseases of PoC. Those pictures will be used to create AI models for different skin types and enhance them over time. For educational purposes doctors and researchers have the possibility to view the image-gallery of classified skin diseases for different skin types.
Currently the design prototype shows the look and feel for our planned product. It shows the workflow of our app for dermatologists. We are further optimizing the amount of features and necessary input data in relation to the time it takes to fill out one dataset.
How we built it
We built this design prototype as a team of two with input from other team members. This is the second version. All work on prototypes cumulates to 50 hours over two people. For our prototype we used the widely known tool figma. This gave us the opportunity to create a design pattern regarding colours and forms and also to implement some animations to create a more realistic feeling. Our prototype will help us as a base for our next step - creating a functional prototype / proof of concept with react.
Challenges we ran into
One challenge was to specify the problem itself as a lot of research is needed and many standards had to be questioned. For example, we had to deviate from initial plans such as setting up an AI model, as increasing research revealed that there are actually not even enough images of people of color for just one particular disease as an experiment. Consequently, we had to change our plans and redefine our basic problem. Since there are many problems to be worked on, it was especially important to find out if similar projects already existed. Initially, we had thought and also hoped that there were at least partial solutions or working groups. However, our market research (see slides) showed that there is actually no similar project with similarly large goals. What we did find, however, is that there has been increased outcry both in research, in the press, and among young physicians in recent years. At the end of the day, nevertheless, there is currently no similar project that specifically accumulates and tests data appropriate for AI models.
Accomplishments that we're proud of
Considering that this is an idea we have been focusing on only since December 2020, we can proudly show our prototype and video presentation. Although our team members are all actively studying and working, we managed to create not only the concept itself and the design prototype, but also our possible business plan provisionally. We are especially proud of the fact that we were able to receive positive and constructive feedback in all the competitions we participated in, and we were even awarded prizes.
- 2. Place - HubIT Online-Designathon with a first prototype and idea outline (Project)
- 1. Place - Common Ground Camp 2021 (award ceremony)
What we learned
Even though we were already aware that there is still a lot of room for improvement in our society, during our research we became more and more aware of the extent to which supposedly objective areas such as medicine and health are also affected. Increasingly, we became aware that even though researched algorithms can have ethically problematic consequences, that some of these problems can be addressed through planning and extra caution. Here, the phrase "insight is the first road to recovery" comes to mind.
In addition to the considerations shown in the presentation, we have also looked at the current market and business aspects. In doing so, we realized how many people could actually benefit from our solution, which is why we resolved to definitely not lose sight of our idea. Some additional information can be found in our slides appendices, where we included a first draft of our business model canvas.
What's next for SKINai 2.0
To describe how things will continue for SKINai, we recommend watching our video presentation. In short:
- We want to use our possibilities to develop a functional prototype with minimal requirements as a proof of concept. On the one hand, the developed design, but above all the imagined functions are to be integrated.
- With this prototype it is possible to ask dermatologists and health experts for their opinion. Therefore it is necessary to work on our network at this point at the latest.
- Several iterations with improvements and continuous testing should be done.
- Once we have enough data for some diseases and skin colors, we want to work on specialized AI models. Before that, the diagnosis option would only be usable with reservations.
Since we are currently still studying and working on the side, this development will take some time.
Built With
- figma
Log in or sign up for Devpost to join the conversation.