This repository contains code for preprocessing Near-Infrared Spectroscopy (NIRS) signals, extracting features, and applying machine learning and deep learning methods to detect adverse outcomes. The primary machine learning method used is XGBoost, and the deep learning approach involves Convolutional Neural Networks (CNNs).
NIRS is a non-invasive optical technique used to monitor tissue oxygenation. This project aims to process NIRS signals to predict adverse outcomes, such as complications during medical procedures. By extracting relevant features from the signals and applying advanced machine learning and deep learning methods, we aim to achieve high accuracy in outcome prediction.
- Signal_processing: Folder containing all the methods to pre-process the NIRS signals and remove the artifacts.
- Feature_extraction: All the methods to extract features from NIRS signals containing transients, and signals without transients and the transients themselves.
- Deep_learning: Source code for CNN models
- utils: Utility functions
For any questions or suggestions, feel free to contact:
Minoo Ashoori - min.ashoori@gmail.com
GitHub: Minimnim