Blink detection algorithm

martinaguas1 · February 11, 2018, 7:57pm

Hi,
I am describing blinks algorithm used by Brainstorm for a paper. I was wondering if, besides the information reported in tutorial #12 “Artifacts detection”, section “Detection: custom events”, do you share more information about the used algorithm (kurtosis?) and methodology?

Thanks in advance,
Miguel.

Francois · February 12, 2018, 2:24pm

Hi Miguel,

This detection method is not published. It is based on the signals’ standard deviation and RMS.
You can read through the code for more details:

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/process/functions/process_evt_detect.m#L297


end
% Return all the input files
OutputFiles = {sInputs(iOk).FileName};
end




%% ===== PERFORM DETECTION =====
% USAGE:      evt = Compute(F, TimeVector, OPTIONS, Fmask)
%             evt = Compute(F, TimeVector, OPTIONS)
%         OPTIONS = Compute()                              : Get the default options structure
function evt = Compute(F, TimeVector, OPTIONS, Fmask)
% Options structure
defOptions = struct('bandpass',     [10, 40], ...   % Filter the signal before performing the detection, [highpass, lowpass]
                    'threshold',    2, ...          % Create an event if the value goes > threshold * standard deviation
                    'blanking',     .5, ...         % No events can be detected during the blanking period
                    'maxcross',     10, ...         % Max number of bounces accepted in one blanking period (to ignore high-frequency oscillations)
                    'ampmin',       0, ...          % Minimum absolute value accepted for a detected peak
                    'isnoisecheck', 1, ...          % If 1, perform a noise quality check on the detected events
                    'noisethresh',  2.5, ...        %    => Noise threshold (x standard deviation or the rms)
                    'isclassify',   1, ...          % If 1, classify the events in different morphological categories
                    'corrval',      .8, ...         %    => Correlation threshold

Cheers,
Francois