Peedicting AHF in CT scans

Prerequisites

1. Download and install uv.
2. Clone the repository git clone https://github.com/orting/breathct.git

Feature extraction

1. Copy your images to data/images. It should be in .nii.gz format.

To segment the data you need a license for the heartchambers_highres model, https://backend.totalsegmentator.com/license-academic/

uv run main.py --feature-extraction

Or to run individual steps

uv run main.py --segment
uv run main.py --merge
uv run main.py --extract

Prediction

1. Copy you patient info to data/patient_info.csv. It must have columns
   • name : Matching the image name without .nii.gz suffix
   • pid : Patient id (One unique pid per patient)
   • age : Age of patient at time of scan

uv run main.py --predict

This will produce data/preds.csv containing a prediction for each pid.