X-Ray image analysis

• 

  Home page

• 

  Image Upload

• 

  Covid positive prediction

• 

  Covid negative prediction

Inspiration

The idea behind the project is to provide a low cost tool that can help diagnosis of diseases using AI and ML in a cost effective manner and with increased accuracy.

What it does

The tool takes in 2D x-ray images as input and classify the image as Covid-19 positive or negative

How I built it

The heart of the project is a machine learning model trained using a dataset of around 120 images ( 60 covid positive and around 60 covid negative cases). The base model used is a VGG16 model on top of which 2 more layers where added to create the model. The web interface is created using flask framework and bootstrap.js The model is deployed to pythonanywhere.com which is a PaaS cloud provider.

Challenges I ran into

To improve the accuracy of the trained model we retrained the model using VGG19 as the base model . But the model file size exploded to 700 MB. Still working on this to identify why this has happened

Accomplishments that I'm proud of

The model deployment to the cloud provider platform which is accessible universally.

What I learned

Transfer learning is one area that i learned during this hackathon. For the earlier models that i have created i had trained the model for the all the layers of the neural network consumes lot of computer resource. By using the transfer learning methodology i was able to reduce the training time for the model drastically,

What's next for X-Ray image analysis

The model can be further enhanced to add more disease prediction from x-ray images. This will require a medical practitioners support to source the images and mark the features. Once the model is tweaked to provide the necessary confidence to medical practitioners it can be production on the web and can be used by anyone in the world for accurately predicting diseases from the x-ray image

Built With

• keras
• python
• tensorflow
• vgg16