When businesses reopen, we need an effective way to test masses. We need a tool to identify COVID-19 symptoms that is cheap, fast, easy to scale and use existing infrastructure. Also after a successful diagnosis, we need to answer additional questions like:
Could this patient stay at home or need an ICU? If a patient needs an ICU, in how many days? What is the survival rate of this patient? How he is going to respond to some particular treatment? Currently, our team is focused on building a diagnosis tool, and we are working on answering the above questions and add these additional features in our tool. In medical AI tools, it's important to be transparent about the model error metrics, data collection, peer review, and clinical study. We will be releasing these details along with our product.
What it does
Our product uses X-rays to identify if it's COVID-19 or pneumonia or normal. Also, it has some explainability to find the reason behind some particular prediction.
How I built it
We used open-source COVID-19 X-rays from Github, and downloaded normal and pneumonia X-rays from Kaggle. Then we built deep learning to model that classifies chest X-rays into any of the following category: Covid-19, Normal, Pneumonia. Then it shows the region that it uses for some particular prediction (AI Explainability)
Challenges I ran into
Implemented various techniques to avoid overfitting the model. Our model uses 998 X-rays, it's hard to train a deep learning model using fewer data. So, we tried to use transfer learning and did various techniques to train the model without overfitting.
Accomplishments that I'm proud of
The model successfully classifies Covid-19 X-rays with 100% recall, and the overall accuracy of the model is 97%. Also, we have a few more ideas to improve the model and overall product experience
What I learned
Product iteration, useability, explainability
What's next for Deep Learning tool for diagnosis and prognosis Covid-19
we need to answer additional questions like:
Could this patient stay at home or need an ICU?, If a patient needs an ICU, in how many days?, What is the survival rate of this patient?, How he is going to respond to some particular treatment?
Also, we are planning to use natural language processing to explain reasons for making some particular prediction