Nepal is one of the poorest countries in Asia, and is home to almost 30 million people. According to Nepal Journals Online in 2017, on average a doctor had to attend 1724 people!!! To make things worse, in the rural areas of Nepal there is on average 1 doctor for 150 thousand people.
We believe, Democratizing healthcare with AI to uplift underprivileged communities.
What it does
In this project, we explore the possibilities of using AI in places with a limited number of experts with exorbitant work and responsibility. We demonstrate how AI may be used to detect abnormalities in muscular radiographs which affects billions worldwide and is the most common cause of severe long-term pain and disability.
How we built it
In order to achieve this, we demonstrate a use-case using the MURA dataset made available by Stanford and build a deep learning model using PyTorch. Specifically, we use Dense-Net (169) for this task and class activation map for interpretability: both using Pytorch.
Challenges we ran into
Hardware resource to train the model and implementing Class Activation Map.
Accomplishments that we're proud of
This hackathon motivated us to conduct research to tackle real-world problems.
What we learned
Research methodology, Pytorch, Implementing Dense-Net architecture.
What's next for Daaktar.AI
Continue the research with the purpose of Democratizing Healthcare.