Archive structure:
data/ Contains some marked EEG segments. Those segments come from the archive http://meg.univ-amu.fr/data_papers/roehri2017/simulation_rate_3.zip. Their simulation process is described in [1]. They are exported to Matlab .mat, files using the AnyWave software (http://meg.univ-amu.fr/wiki/AnyWave).
Note that on Linux, to run AnyWave from its folder, I need to use:
`env QT_PLUGIN_PATH=plugins/ ./AnyWaveLinux`
demo_matlab/ and demo_octave/ Contain the “proof of concept” detector described in the report, respectively optimised for Matlab and for Octave (the main differences are in the .mex and .oct files), the scripts are mostly the same thanks to Octave being compatible with the syntax of Matlab. From a Matlab or an Octave instance having the Time-Frequency Toolbox (http://tftb.nongnu.org/) avalaible on its path, just run:
>> detector_demo
stats_matlab/ and stats_octave/ Contain a script that produces statistics on a number of marked HFOs. Same remark as for demo_{matlab,octave} regarding the diffrences between the two folders. Here run:
>> detector_stats
If needs be, one can recompile the native code using either:
from Matlab:
>> mex CXXFLAGS='\$CXXFLAGS -std=c++11" find_components_mex.cpp
>> mex find_zeros_mex.cpp
of from Octave:
> mkoctfile find_components_oct.cpp
> mkoctfile find_zeros_oct.cpp
Bibliography: - [1] : Roehri, N., Pizzo, F., Bartolomei, F., Wendling, F., & Bénar, C. G. (2017) What are the assets and weaknesses of HFO detectors? A benchmark framework based on realistic simulations. PloS one, 12(4), e0174702.