Inspiration
It usually takes days for doctors to diagnose a patient's medical images accurately this poses a time constraint on both the doctor and the patient (has to wait for a diagnosis before starting treatment) using a CNN that has been trained on large datasets doctors can quickly diagnose complex cases with a high degree of confidence.
What it does
This idea uses machine learning to diagnose pneumonia in patients. The model accepts patient chest X-rays and after analysis can predict with 94% accuracy whether the patient has pneumonia or not.
How we built it
We trained a Convolutional Neural Network using a patient X-RAY dataset that we found on Kaggle. When fed with new X-ray images this model could predict with an accuracy of 94% whether the patient was suffering from Pneumonia or not. To add another layer of complexity to the project we researched common pneumonia symptoms and gave them a weighting based on how impactful they were in terms of determining if a patient had pneumonia or not. An equation was created that used these symptoms and generated the probability of the patient having pneumonia.
Challenges we ran into
The main challenge we faced was in connecting the CNN to the back end of our web application. This was the first time our team members were using an AI model coupled with a web application and the learning curve proved to be steep.
Accomplishments that we're proud of
Our idea will allow doctors to confirm/predict a patient's diagnosis in a short period such that what would normally take days can now be done in mere seconds.
This was our first experience of using AI with web applications. The experiences and challenges we faced will surely help us in the future, especially with the boom in AI-based applications.
What we learned
Learned how to create a fully-fledged web application that implements a machine-learning model. Learned new technologies like flask and TensorFlow.
TRAINING OUR MODEL
https://www.kaggle.com/code/madz2000/pneumonia-detection-using-cnn-92-6-accuracy
DISPLAYING OUR UI
https://mui.com/material-ui/getting-started/installation/
What's next for SmartDiagPro.
In the future, this application will be extended for other diseases that require medical imaging. The application will essentially act as a doctor's assistant.
Log in or sign up for Devpost to join the conversation.