as deep learning in general and computer vision in particular has advanced in recent years. and process of training computer to understand and interpret images. Machines can accurately identify and classify objects. deep learning starts to evolve in healthcare sector and improve a lot of processes, therefore i thought about creating wepapp that can help doctor diagnose lymphoma cancer, or even using it as learning tool
What it does
it classifies between Follicular lymphoma (FL), Chronic lymphocytic leukemia (CLL), and Mantle Cell lymphoma cancer by using Microscopic image from patients tissue
How I built it
first find the right dataset, later i used tensorflow2.0 api to build CNN to classify three types of lymphoma cancer. finally i convert trained model to be served in tensorflowjs in browser , and build a game using angular, where computer using trained model to try to identify what type of lymphoma cancer it is.
Challenges I ran into
- preprocess data, and find the right layers structure for CNN to avoid overfitting
- reduce trained model size to fit on browser
Accomplishments that I'm proud of
understand how Convolutional Neural Network works
What I learned
i learned a lot about Convolutional Neural Network (CNN), Tensorflow 2.0 Api and how to serve trained model offline in browser
What's next for Computatrum
need to collect more data to re-train model and improve accuracy and decrease over-fitting.