Summer school: Human Interfacing and Augmentation. Held in London, from june 30th to july 4th 2025.
Webinar: Customer discovery and Market Analysis. Held Online, March 27th 2025.
HybridNeuro Hub Registration.

Short example of Biceps Brachii muscle surface HDEMG decomposition using the DEMUSE Tool

  • Author: Aleš Holobar

Keywords: surface high density electromyogram (HDEMG), decomposition, motor unit, DEMUSE, simulation, biceps brachii, dataset

SHARING/ACCESS INFORMATION
This dataset is available under the Creative Commons Public Domain Dedication (CC0) license.

Publicly accessible location of the data: https://zenodo.org/records/10936952

DOI: 10.5281/zenodo.10936952

Recommended citation for this dataset:
Holobar, Aleš. “Short Example of Biceps Brachii Muscle Surface HDEMG Decomposition Using the DEMUSE Tool”. Zenodo, April 6, 2024. https://doi.org/10.5281/zenodo.10936952.

DATA & FILE OVERVIEW
This dataset contains 4 examples of synthetic high density surface EMG signals of the Biceps Brachii muscle and results of their decomposition into separate motor unit activity. It is intended as a demonstration of the DEMUSE Tool software for sEMG decomposition and as a basis for practical example of dataset preparation for the HybridNeuro project webinar on Data management and ethics (https://www.hybridneuro.feri.um.si/results.html#webinars ). Two sets of data are included: the raw simulated sEMG signals and the results of decomposition of those signals with the DEMUSE Tool.

File list:
  • README.txt: this text file containing description of the dataset
  • metadata.xml: Dublin Core metadata in XML format
  • folder "raw_signals":
  •      - sEMG_SynthSig_BicepsBrachii_F10MVC_Len20_SNRInf_03-Apr-2024_rawSignals.mat: simulated raw sEMG at force level 10
  •      - sEMG_SynthSig_BicepsBrachii_F30MVC_Len20_SNRInf_03-Apr-2024_rawSignals.mat: simulated raw sEMG at force level 30
  •      - sEMG_SynthSig_BicepsBrachii_F50MVC_Len20_SNRInf_03-Apr-2024_rawSignals.mat: simulated raw sEMG at force level 50
  •      - sEMG_SynthSig_BicepsBrachii_F70MVC_Len20_SNRInf_03-Apr-2024_rawSignals.mat: simulated raw sEMG at force level 70
  •      - MAT_file_variables_rawSignals.png: screenshot of variables from raw signal files loaded into Matlab
  • folder "decomposed":
  •      - sEMG_SynthSig_BicepsBrachii_F10MVC_Len20_SNRInf_03-Apr-2024_DEMUSE_edited.mat: decomposed and manually cleaned simulated sEMG at force level 10
  •      - sEMG_SynthSig_BicepsBrachii_F10MVC_Len20_SNRInf_03-Apr-2024_DEMUSE_edited.mat: decomposed and manually cleaned simulated sEMG at force level 10
  •      - sEMG_SynthSig_BicepsBrachii_F30MVC_Len20_SNRInf_03-Apr-2024_DEMUSE_edited.mat: decomposed and manually cleaned simulated sEMG at force level 30
  •      - sEMG_SynthSig_BicepsBrachii_F50MVC_Len20_SNRInf_03-Apr-2024_DEMUSE_edited.mat: decomposed and manually cleaned simulated sEMG at force level 50
  •      - sEMG_SynthSig_BicepsBrachii_F70MVC_Len20_SNRInf_03-Apr-2024_DEMUSE_edited.mat: decomposed and manually cleaned simulated sEMG at force level 70
  •      - MAT_file_variables_DEMUSE_edited.png: screenshot of variables from DEMUSE result files loaded into Matlab
  •      - DEMUSE_Tool_parameters.png: screenshot of the DEMUSE Tool with the selected parameters

All files with extension .MAT are in Matlab format, which is a proprietary format but its specifications are open and there are open source routines available for reading and writing:

  • The GNU Octave software, which is an open source alternative to Matlab, can read and write .MAT files.
  • The Python library Scipy can load MAT files.
  • The Matio project on SourceForge is a C library for reading and writing Matlab MAT files.
  • MAT files with suffix "_rawSignal" contain the raw, unedited HDEMG signals (+ related information).
  • MAT files with suffix "_DEMUSE_edited" contain results of HDEMG decomposition into separate motor units using the DEMUSE Tool software (version 6.3).

METHODOLOGICAL INFORMATION
Description of methods used for collection/generation of data:
The synthetic multichannel sEMG data were generated by a multilayer cylindrical volume conductor model [1] for the biceps brachii muscle.

Model parameters:
  • Motor unit conduction velocity: 4 m/s
  • Tissue conductivities (in S/m):
  •      - Bone: 0.02
  •      - Muscle (radial and transverse): 0.1
  •      - Muscle (longitudinal): 0.5
  •      - Subcutaneous tissue: 0.05
  •      - Skin: 0.1
  • Muscle properties:
  •      - Number of motor units: 500
  •      - Number of muscle fibers: 165656
  •      - Number of fibers in a motor unit (range): 25 - 1500
  •      - Muscle cross-sectional area: 1413 square mm
  •      - Average fiber length: 130 mm
  •      - Skin thickness: 1 mm
  •      - Subcutaneous tissue thickness: 4 mm
  •      - Bone (radius): 20 mm
  •      - Tendon ending spread: 5 mm
  • Electrodes:
  •      - Grid: 10 rows x 9 columns
  •      - Circular (diameter): 0.5 mm
  •      - Interelectrode distance: 5 mm
  • Excitation level (% maximum):
  •      - Constant force: 10, 30, 50, 70% of maximum voluntary contraction (MVC)
  • Innervation zone in the middle
  • Sample rate: 4096 Hz
  • Total length: 20 s
  • SNR: infinite (no noise added)

Motor unit (MU) firing pattern model: The MU firing pattern was computed by the model proposed in [2] with the parameters adapted to the biceps brachii muscle. MU recruitment thresholds followed an exponential distribution with many low-threshold MUs and progressively fewer high-threshold MUs [2]. The last MU was recruited at 80% of maximal excitation level. MUs linearly increased the firing rates from 8 pulses per second (pps) at recruitment to 35 pps at 100 % of maximal excitation. Interspike interval of each MU followed Gaussian distribution with coefficient of variation equal to 20 %.

Methods for processing the data:
The raw multichannel sEMG signals generated by the simulator were downsampled to 2048 Hz (which is closer to sampling rate typically used in experimental recordings) and are available in the .MAT files with filename suffix "_rawSignal". The raw sEMG signals were loaded into the DEMUSE Tool software (https://demuse.feri.um.si/), which was used to decompose the signals into a set of motor units and their firing instants.

DEMUSE Tool parameters:

  • Band-pass filter: 20 - 500 Hz
  • High-pass differential filter: on
  • Spatial filter: none
  • Notch filter for removing line interference: off
  • Estimated percentage of good quality EMG channels: 95%
  • Automatic channel selection: off
  • Manual channel selection: on
  • Number of decomposition runs: 50
  • Offset of the EMG signal epoch: 0 s
  • Length of the EMG signal epoch: 30 s
  • Decompose signals in non-overlapping mode: off

Quality-assurance procedures performed on the data: The results of decomposition were manually inspected by an expert and any motor units that had pulse-to-noise ratio (PNR) lower than 28 dB [3] were discarded in order to ensure high accuracy of motor unit firing identification.

DATA-SPECIFIC INFORMATION FOR .MAT FILES CONTAINING RAW EMG SIGNALS ("_rawSignals" suffix)
Number of variables: 7

Variable list:
  • fsamp: integer number. Data sampling frequency in Hz.
  • SigLength: double number. Total length of the signals in seconds.
  • sig_out: structure array. Dimensions are the same as the simulated EMG electrode grid: 10 rows x 9 columns. Each cell contains a 1 x N vector of samples (double values) representing EMG voltage. N = total number of samples = fsamp * SigLength
  • sFirings: structure array. Contains 500 cells, each cell contains a 1 x M vector of motor unit firing instants (integer values), expressed as sample indices of the sig_out vectors. M = number of motor unit firings
  • MUAPs: structure array. Contains 500 cells, each cell has 9 sub-cells, each sub-cell contains 10 x 128 matrix of double values, each of the 10 rows contains one motor unit action potential (MUAP) represented with 128 samples.
  • RecThresh: double number. Threshold value for motor unit recruitment.
  • Force: double number. The exerted muscle force, as percentage of the Maximum Voluntary Force.

DATA-SPECIFIC INFORMATION FOR .MAT FILES CONTAINING DECOMPOSITION RESULTS ("_DEMUSE_edited" suffix)
Number of variables: 21

Variable list:
  • description: string. A short text description containing the main parameter values.
  • fsamp: integer number. Data sampling frequency in Hz.
  • SigLength: double number. The total length of the signals in seconds.
  • MUPulses: structure array. Contains 1 x L cells. Each cell contains a vector of detected motor unit dicharge instants (integer values), expressed as sample indices of the SIG vectors. L = number of detected motor units.
  • MUIDs: structure array. Contains 1 x L cells. Each cell contains a string with the motor unit label. L = number of detected motor units.
  • IPTs: 2D array of double values with L rows x N columns. Contains innervation pulse train (IPT) values as estimated by DEMUSE decomposition tool. L = number of detected motor units. N = total number of samples = fsamp * SigLength.
  • PNR: vector of double values with L elements. Contains calculated pulse-to-noise ratio for each detected motor unit. L = number of detected motor units (see [3] for details).
  • SIGFileName: string. Name of the file containing the input EMG signals for decomposition.
  • SIGFilePath: string. Path of the file containing the input EMG signals for decomposition.
  • startSIGInt: double number. Start time of the data interval used for decomposition.
  • stopSIGInt: double number. End time of the data interval used for decomposition.
  • SIG: structure array. Contains a copy of the input EMG signals. Dimensions are the same as the simulated EMG electrode grid: 10 rows x 9 columns. Each cell contains a 1 x N vector of samples (double values) representing output EMG voltage. N = total number of samples = fsamp * SigLength.
  • ref_signal: vector of double values with N elements. Contains the exerted muscle force values used as a reference, expressed as percentage of the Maximum Voluntary Force.
  • discardChannelsVec: 2D array of integer values, with the same dimensions as the simulated EMG electrode grid: 10 rows x 9 columns. Contains value 1 if the related electrode channel was discarded from analysis, or value 0 otherwise.
  • AUXchannels: auxilary channel information , unused.
  • Cost: structure array. Contains 1 x L cells. Each cell contains cost (double value) of cost function, used by DEMUSE tool to estimate the regularity of MU firing pattern (i.e. coefficient of variation of interfiring interval).
  • origRecMode: string. Label for the recording mode of the original EMG signals.
  • DecompRuns: integer number. Number of DEMUSE Tool decomposition runs performed.
  • DecompStat: empty, unused.
  • IED: integer number. Interelectrode distance in mm.
  • ProcTime: integer number. Processing time of the decomposition, unused.

REFERENCES:
[1] Farina, Dario, Luca Mesin, Simone Martina, and Roberto Merletti. "A surface EMG generation model with multilayer cylindrical description of the volume conductor." IEEE Transactions on Biomedical Engineering 51, no. 3 (2004): 415-426. doi: 10.1109/TBME.2003.820998.

[2] Fuglevand, Andrew J., David A. Winter, and Aftab E. Patla. "Models of recruitment and rate coding organization in motor-unit pools." Journal of neurophysiology 70, no. 6 (1993): 2470-2488.

[3] Holobar, Aleš, Marco Alessandro Minetto, and Dario Farina. "Accurate identification of motor unit discharge patterns from high-density surface EMG and validation with a novel signal-based performance metric." Journal of neural engineering 11, no. 1 (2014): 016008.