Inspiration

Some of our closest family members were diagnosed with or victimized by cancer. Given the deadly nature of cancer, early testing of cancer becomes extremely crucial. But the current tests which are based on CT and MRI are invasive, expensive, and time-taking. Often, these tests also result in false positives, which leads to unnecessary stress for the individual and their family. This resulted in us asking many questions about cancer testing and this burning desire to obtain a better cancer-detecting method inspired us to look into the current research and develop an algorithm that would accurately detect several types of cancer, all with the help of a simple blood test.

What it does

AccurateDetect is prediction software focusing on the development of a web-based multi-platform solution for augmenting prognostic strategies to diagnose different types of cancer. We use a neural network with 2 hidden layers. The NN takes several parameters from the blood test as input and informs the user if he/she has a particular type of cancer. AccurateDetect was evaluated using ten-fold cross-validation and finally optimized by adjusting hyperparameters and the output threshold. The project achieves a specificity of above 99% and a median accuracy of 82.30%.

How I built it

We built AccurateDetect with Python, Ruby-on-Rails, HTML, CSS, and Javascript. In addition, we used Flask to integrate it with a web app that will allow easier accessibility. The deep learning and machine learning aspects of AccurateDetect were built with Tensorflow and Keras primarily.

Challenges I ran into

My teammate and I didn't have a lot of web development experience. We couldn't host the web app on GitHub unfortunately. We couldn't add more textures and features to the web app because of time constraints. Since we were only 2 people working in it, we couldn't explore the APIs of other services like Auth0, Twilio, etc. We also had plans of making an Android app for the same, but couldn't because of limited team members and time constraints But despite our challenges, we enjoyed the opportunity and are grateful for that.

Accomplishments that I'm proud of

We learned how to create a really user-friendly web app in less than 13 hours. We are proud of our research as we extensively into peer-reviewed journal research, and developed techniques to incorporate these ideas in our project. We also take pride that our project is just a step forward to revolutionize cancer testing.

What I learned

We deepened our understanding of AI and ML, and we also greatly improved upon our WebDev and Python experience. We built upon our research skills, and also learned that the most productive time to code is very late at night :)

What's next for Accurate Detect

Currently, we are very happy with the results of the ML model and are actively looking into getting it in the hands of researchers to deploy into the real world and help get regulatory approval. We are also working on tweaking our neural network to give more accuracy to our model. We also aim to collect more training data for our model from trusted sources. We are looking forward to fully implement and develop our idea to completion, even after the hackathon ends.

Built With

Share this project:

Updates