Calcifer
Artificial Intelligence-based Medical Diagnosis Platform
This project was developed as a hackathon entry for WhiteCloak's Artificial Intelligence Hackathon 2021. I led a team of 3 Software Engineers and 1 UI/UX Designer to build a Web Platform that uses Artificial Intelligence in the form of a Computer Vision Model to potentially detect medical conditions like:
- Malignant Skin Lesions (Cancerous Skin Lesions)
- Pneumonia - From Lung X-ray Images
As I was designing this project, I wanted to emphasize having a friendly user experience that comes in the form of smooth and colorful animations inspired by the Reflecty mobile app. The logo and the name for the platform were inspired by the Studio Ghibli Film: Howl's Moving Castle, in which Calcifer is a living flame entity that changes color and controls the Castle.
I thought that it went well with the idea that the app changes color when it has a bad result or detection coming from the AI Model. The project was built with NextJS and the AI models were trained using datasets from Kaggle. The Teachable Machine platform was used to build the models for the different detection modes.
The final Model we created was an Image Classification model. Though it is simple by design, we were able to implement the following functionalities:
- Identify Invalid input data
- Suggest nearby doctors based on findings
And after many tests, we have a model reaching 93% of accuracy. This was great for a prototype, but we did emphasize in the results, that digital platforms for medical applications still require further review and the accuracy of the model has to be further scrutinized by medical specialists.
At the end of the Hackathon, our team won the grand prize. But more importantly, this project was a great experience for me and my team. This project taught us many things and allowed us to build something useful with technology.
