Shepherd

Shepherd is a hospital AI that helps with managing and monitoring patients, predicting diagnostics, correctly identifying brain tumors and tracking cell growth, and assists with surgeries.

Comment

Inspiration

Brain tumors are one of the most dangerous diseases because they frequently go undetected until it is too late. Doctors and specialists read the brain scans themselves, which is liable to human error. Our AI's goal is to read the scans to catch tumors that the doctors may have missed.

What it does

Shepherd uses current data from the vital readings of the patient and previous medical data from a hospital's patient records API to accurately predict the diagnostic. Given an MRI scan of a brain, the AI can confidently identify a tumor and tracks cancerous cell growth. This AI helps with scheduling operations and tests for the patients. It also helps with surgeries in the operating room.

How we built it

Using Adobe XD, we designed the app for iOS. Since tablets are used more often in hospitals and our team doesn't have an iOS device, we rented an iPad from our university's technology center. Using Java, we call our APIs for patient records and create our user interface. With python and machine learning, we created a CNN model that classifies tumors and tracks cancerous cell growth. Using Google Cloud technologies, we were able to keep track of the APIs and data from our database. Google Cloud also offers us technologies to help with machine learning and AI.

Challenges we ran into

This was our first time developing for iOS. Creating the predictions of diagnostics was difficult since there was so much data we needed to use to correctly diagnose patients. The AI for detecting tumors was also giving us problems. We wanted a higher level of confidence from the tests so we added more images to train the AI to be more accurate by using multiple APIs and databases. Also training the voice commands for surgery was difficult too.

Accomplishments that we're proud of

We are proud of the iOS development we have created. The CNN module was also a very large set of data that we could tackle.

What we learned

We learned how to make an iOS app, machine learning with python, and data structures using Java.

Share this project:

Updates