UCU Software Architecture for Data Science in Python Course 2021
Our project aims to set up a server to perform lesion detection/segmentation on CT scans. The lesion detection task is highly time-consuming and costly. Radiologists have to spend a lot of time manually adding bookmarks. What is more, they usually add it only for one lesion's key slide per CT scan. We, however, would like to optimize this process and provide anyone with an opportunity to upload their CT scan and use our model to detect/segment a lesion on one's scan.
Here we develop a prototype that works with DeepLesion Dataset, and in the future project can be adapted to other datasets.
Step 1: Simply set up of a server
Option 1: Run server locally
pyenv install 3.9.2
pyenv virtualenv 3.9.2 deep-lesion
source <path-to-env>/deep-lesion/bin/activate
python -m pip install --upgrade pip
pip install wheel
pip install -r requirements.txt
uvicorn setup:app --host 0.0.0.0 --port 8080Option 2: Run in docker container
docker build . -t deep-lesion:latest
docker run -p 8080:8080 deep-lesionOption 3: Deploy the app
- Draft a new release using workflow
- Wait for a successful build and Google Cloud Registry upload
- Find your image in GCR with the tag used during the build
- Configure to run on the cloud
Current url
Step 2: Upload your CT scan image (from DeepLesion dataset) as a *.png, *.jpeg or *.jpg file
Step 3: Get detection/segmentation results
