PyTorch 2 Model for the prediction of single-artery segmentations for "AngioPy segmentation: An open-source, user-guided deep learning tool for coronary artery segmentation"

These are pytorch model weights for the angiopy-segmentation deep learning model, descriped in a paper published in 2024 in the International Journal of Cardiology: AngioPy segmentation: An open-source, user-guided deep learning tool for coronary artery segmentation.

These help to predict a binary mask for a single artery based on a single time-step of a greyscale angiography image series, given a few clicked points which are provided to the model as "clicked pixels".

This model was trained based on manual pixel-level annotations by Dr. Mahendiran (Cardiologist in CHUV, Lausanne, Switzerland) during an academic visit to EPFL, Lausanne, Switzerland on anonymised image set coming from Invasive Coronary Angiographies selected at the time step closest to end-diastole in the cardial cycle. Angiographies were all from the Fractional Flow Reserve versus Angiography for Multivessel Evaluation 2 (FAME 2) trial.

Details of the trial have been published previously [10.1056/NEJMoa1205361]. The trial conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the participant institution's human research committee.