Submitted for: Most Socially Useful Hack
See more videos:
It is utmost important to do regular medical check! But we want the doctors to be less involved!
In this project, we managed to build a artificially intelligent diagnosis systems
We built an iPhone app to be your intelligent personal doctor who is somewhat both cardiologist and dermatologist. The app captures 10x zoomed image of human skin nevus or any suspicious region to detect skin cancer.
Problems to solve
1 - Skin Diseases Detection (iPhone App)
What it is
Our deep neural network model trained on a skin dataset achieved around 84% accuracy to detect the presence of skin diseases.
What it does
The app will capture the image and send it back to a server for diagnosis and return the result back, along with some recommendation and advice. It will identify if the skin is healthy or ill.
How we built it
We train a neural network to train on �dataset and deploy it to in Python Flask. The iPhone app will send the image to this server and retrieve the diagnosis results
2 - Chest X-Ray Diagnosis for Lung and Heart Disease (webapp - concept)
What it is?
There are many different diseases that can be discovered through Chest X-ray Imaging. In this project, we try to diagnose 14 different lung and heart-related diseases and defects.
We trained our machine learning model on a dataset and reached around 90% accuracy. Note that non-experts may not be able to identify these problems when looking at the image.
What it does
We demonstrate a concept of portable X-Ray Imaging Booth (like a PhotoID booth) using a web app. The application will show yourself on the webcam and instruct you to take an x-ray image in the X-Ray booth. It will instantly identify the probabilities of 14 different diseases and defect in the chest.
How we built it
We built a deep learning model and trained it on the dataset and use it to serve inference on the computer. We built a simple website to allow uploading x-ray image for real-time diagnosis.
Challenges I ran into
This is the first we develop iOS and swift application. So it was a little bit hard. The dataset we used is imbalanced and it was difficult to train.
Accomplishments that I'm proud of
We managed to develop our first Swift iOS app, which makes us happy. We are proud that we managed to solve difficulties in medicine and bring a better health-care solution to the communities
What I learned
We learn a lot about machine learning, deep learning, different kinds of neural networks. We also learn to code iOS, which we're never done before. We also learn about general medical knowledge.
What's next for Deep Doctor
- We want to further develop it to support more kinds of diseases, improve the accuracy.
- We also want it to be able to provide a correct treatment plan and medicines.
- We need to collect a lot more data in order to accomplish that.