CancerScan

• 

Inspiration

As we know brain cancer is a very menacing and invisible problem. Thus it becomes imperative to know early whether we have developed it or not. Our tool is mainly targeted in regions where there is a shortage of expert radiologists and people can get an idea whether they might have brain cancer or not.

What it does

Cancer Scan is an AI-powered web tool that classifies brain MRI images into categories like:

1) No tumor

2) Pituitary tumor

3) Meningioma tumor

4) Glioma tumor

Users can upload an MRI image through our user-friendly interface, and our trained deep learning model instantly analyzes the scan and provides a prediction with a confidence score. The goal is to provide everyday users with a fast, lightweight tool for assisting in brain tumor detection before they consult a doctor

How we built it

We developed CancerScan using the following technologies:

+~ Frontend - HTML, CSS, JavaScript

+~ AI Model - Transfer learning using RoboFlow API (VIT classification model)

+~ Hosting - Vercel

+~ Collaboration - Github

Challenges we ran into and how we overcame them

1) API Errors (400/404) - Reviewing the Roboflow API documentation, and debugging through console logs

2) Roboflow Public Key Limitations - Switched to private key

3) Frontend Debugging - DevTools and console logs to inspect network requests

Accomplishments that we're proud of

1) Successfully training and testing a brain MRI classifier with high accuracy across multiple tumor classes.

2) Creating a clean, functional web interface for MRI uploads and predictions.

3) Establishing strong team collaboration with consistent late night google meet sessions.

What we learned

1) How to use the Roboflow API

2) Debugging through our browser's console

2) Collaboration on a multi disciplinary project

What's next for CancerScan

We plan to have many iterations including the detection of a wider range of brain cancers and other types of diseases.

Built With

• css
• html
• javascript
• roboflow