spindles-HMM

Code to replicate all results, tables, and figures of paper "Robust autoregressive hidden semi–Markov models applied to EEG sleep spindles detection"

Requirements

MATLAB - Results, Tables and Figure 2

• MATLAB R2016b or newer (the implemented algorithms require broadcasting capabilities)
• Statistics and Machine Learning Toolbox
• Signal Processing Toolbox
• Parallel Computing Toolbox (optional, but extremely recommended for EM-based learning)

Python - Figure 1

• python 3
• jupyter notebook
• matplotlib
• daft to plot the probabilistic graphical model

Data

• DREAMS sleep spindles dataset

Instructions

1. Add "Main_Code" folder to MATLAB path
2. Download DREAMS sleep spindles dataset. A reliable source is here (make sure to download the DatabaseSpindles.rar file). Unpack .rar file.
3. Run reformatDREAMS.m script to reformat and downsample the EEG and expert labels
4. Algorithms/scripts in the "DREAMS" folder are now ready to be run. These are the scripts that replicate results, tables, and Figure 2
5. (Optional) Set up a python environment with matplotlib and daft to run RARHSMM_GraphicalModel.ipynb on a jupyter notebook

PS1: All main code is MATLAB, python is only used to render the probabilistic graphical model of Figure 1

PS2: The DREAMS sleep spindles dataset used to be hosted here but that website has been down for quite some time now