Approximate Entropy and Absolute Band Power -Frequency Bands

Feature extraction of nonlinear (approximate entropy) and linear (absolute band power) band features for disease characterization

Pograms used for feature extraction in neuropathic pain characterization (n=36) vs a control group (n=13). Thus, three sets of codes are available.

Main.m -> main program that applies function of FeatureExtraction to each subject (if subjects are in ID folders)

FeatureExtraction_NP.m -> Function that extracts approximate entropy and absolute band power for 22 electrodes in the five clinical frequency bands (delta, theta, alpha, beta, gamma)

ENTROPY.m -> Function that applies approximate entropy algorithm using two functions of the Predictive Maintaeance Toolbox: PhaseSpaceReconstruction and approximateEntropy. Dimension (m) is defined as 3, but the lag or "delay" is computed specifically for subject and condition.

The same programs but applied for a control dataset included feature extraction of 19 electrodes and 13 subjects:

Main_control.m

FeatureExtraction_control.m

ENTROPY.m (the same)

% ====== FEATURE EXTRACTION For Chronic NP - 36 subjects======

% (1) Source 1: Entropies = 5 x 22

% -- 5 bands: delta, theta, alpha, beta, gamma

% -- 22 channels

% (2) Source 2: Absolute Band Power = 5 x 22

% -- 5 bands: delta, theta, alpha, beta and gamma

% -- 22 channels

% (3) Total number of features: 110 x 110 = 220 (K)

% (4) Total number of observations: 10 segments x 36 participants = 360 (N)

% (5) Apply FeatureExtraction function for the individual vector (consisting of Algorithm 1 and 2)

% (6) Both entropy and power features (220 K) are returned to the original features vector in Main program, and next subject is computed