The implementation of surgicalVAE (surgical Variational Autoencoder),
the innovative and highly accurate VAE-based Perioperative Prediction Framework (Semi-supervised and Multi-tasking model),
capable of simultaneous prediction across N = 6 different crucial outcomes in high-risk cardiac surgery.
The repo contains training and testing codes.
Trained model files can be downloaded from here: https://figshare.com/s/e44f0120502b01583ba2. Trained under 5-fold cross-validation settings specified in the paper.