Data in Brief 53 (2024) 110250

Contents lists available at ScienceDirect

Data in Brief

journal homepage: www.elsevier.com/locate/dib

Data Article

A multichannel electromyography dataset for

continuous intraoperative neurophysiological
monitoring of cranial nerve
Wanting Ma a, Lin Chen a, Xiaofan Pang b, Yuanwen Zou a,∗
a
College of Biomedical Engineering, Sichuan University, 610065, China
b
Department of Neurosurgery, West China Hospital, Sichuan University, 610041, China

a r t i c l e i n f o a b s t r a c t

Article history: Continuous Intraoperative Neurophysiologic Monitoring

Received 18 December 2023 (cIONM) is a widely used technology to improve surgical
Revised 16 February 2024 outcomes and prevent cranial nerve injury during skull base
Accepted 20 February 2024 surgery. Monitoring of free-running electromyogram (EMG)
Available online 27 February 2024
plays an important role in cIONM, which can be used to
Dataset link: A Multichannel Continuous
identify different discharge patterns, alert the surgeon to
Clinical Electromyography Dataset from potential nerve damage promptly, etc. In this dataset, we
Neurosurgery (Original data) collected clinical multichannel EMG signals from 11 inde-
pendent patients’ data using a Neuromaster G1 MEE-20 0 0
Keywords:
system (Nihon Kohden, Inc., Tokyo, Japan). Through innova-
Intraoperative Monitoring
tive classification methods, these signals were categorized
Skull base surgery
Electromyography (EMG)
into seven different categories. Remarkably, channel 1 and
Face nerve channel 2 captured continuous EMG signals from the facial
Deep learning nerve (VII cranial nerve), while channel 3 to channel 6
focused on V, XI, X, and XII cranial nerves. This is the first
time that intraoperative EMG signals have been collated
and presented as a dataset and labelled by professional
neurophysiologists. These data can be utilized to develop the
architecture of neural networks in deep learning, machine
learning, pattern recognition, and other commonly employed
biomedical engineering research methods, thereby providing
valuable information to enhance the safety and efficacy of
surgical procedures.

∗
Corresponding author.
E-mail address: [email protected] (Y. Zou).

https://doi.org/10.1016/j.dib.2024.110250
2352-3409/© 2024 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC license
(http://creativecommons.org/licenses/by-nc/4.0/)
2 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250

© 2024 The Authors. Published by Elsevier Inc.

This is an open access article under the CC BY-NC license
(http://creativecommons.org/licenses/by-nc/4.0/)

Specifications Table

Subject Neurosurgery / Neurophysiology

Specific subject area Biomedical Signal Processing, Clinical Electromyography, Intraoperative
monitoring, Electrophysiology, EMG Classification
Data format Raw
Type of data Tables
Data collection We collected raw data utilizing the Neuromaster G1 MEE-20 0 0 (Nihon Kohden,
Inc., Tokyo, Japan) using free-running EMG, TcMEP from 11 participants. Paired
subdermal needle electrodes (1500/13/0.4 mm) were inserted in the special
areas. A filter was set between 5 and 2 kHz. After the surgery, we exported
EMG data and an event list (labels) via the Neuromaster Review Mode V05-11
(Nihon Kohden, Inc., Tokyo, Japan) software. Due to the specificity of the
clinical data, no special filtering or preprocessing was performed on the data
so as to avoid distorting special categories of signals and to preserve their
authenticity.
Data source location Operating Room, department of neurosurgery, West China Hospital of Sichuan
University, No.37, Guoxue Lane, Wuhou District, Chengdu, Sichuan, China.
Data accessibility Repository name: Mendeley
Data identification number: 10.17632/7hyptcbkkd.2
Direct URL to data: https://data.mendeley.com/datasets/7hyptcbkkd/2

1. Value of the Data

• It’s the first dataset that contains six-channel continuous clinical EMG data and also clas-
sifies and labels a-train, b-train and other classic cranial nerve injury patterns, which is
valuable in neurophysiology, intraoperative cranial nerve preservation, biomedical engi-
neering, pattern recognition, deep learning, etc.
• These data were collected from real neurosurgical surgeries, which are often very difficult
to obtain and require a great deal of time and effort to collate.
• We collected a large amount of intraoperative continuous EMG data and spent a lot of
time on expert annotation by neurophysiologists. The data are more generalizable and
clinically meaningful compared to the previous guided, fixed-posture EMG data performed
in the laboratory.
• This dataset is valuable for researchers engaged in the field of EMG classifications, pattern
recognition, the recognition of specific EMG signals corresponding to cranial nerve injury
and developing new automatic warning methods of intraoperative monitoring.
• In the field of intraoperative cranial nerve preservation, manual identification of monitor-
ing signals is required during surgeries. It leads to a high degree of subjectivity in intraop-
erative monitoring and alarm criteria, making it difficult to have a more uniform standard.
This dataset is proposed to help solve this problem. In addition, it can also provide useful
information for medical practices to improve the preservation of cranial nerves and the
safety of surgeries.
• In the field of biomedical engineering, it can help to solve the problem of transient abnor-
mal signals that may be missed by insufficient manpower. This dataset has the potential to
inspire researchers to develop real-time alarm systems for clinical surgeries. This dataset
can be used to develop more deep-learning-based or machine-learning-based algorithms,
models, or tools for monitoring neurological function related to surgeries and to study
the mechanisms and effects of nerve injury or repair during surgery. Other applications in
biomedical engineering are also valuable such as explore the relationship between A-train
 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250 3

Table 1
The average profile of patients participating in the dataset.

Gender Age (Mean±SD) Tumor site

Male (n = 5) Female (n = 6) 49.73 ± 16.65 Left (n = 5) Right (n = 6)

discharge pattern and postoperative facial paralysis or automated elimination of interfer-

ence from similar signals that are not easily distinguishable.

2. Background

Cranial nerve injury is a common complication following skull base surgery [1], and can lead
to various adverse effects. Nowadays, continuous Intraoperative Neurophysiologic Monitoring
(cIONM) has been introduced to minimize the risk of such injuries and aid in the protection of
nerves during surgery by continuously monitoring and recording the functional status of specific
nerves [2]. It has become one of the most commonly used methods in neurosurgery to enhance
surgical outcomes and prevent cranial nerve injuries [3]. However, there are some limitations
associated with the use of cIONM to monitor free-running electromyography (EMG) activity in
real-time, making it difficult to establish consistent monitoring and alarm standards. The robust-
ness of the signal recognition and the subjectivity of the alarms are yet to be solved. Although
relevant studies [4,5] found that some special EMG discharge patterns are highly correlated with
postoperative cranial nerve palsy [6], further research is still needed. To help more researchers
solve these problems, we compiled this six-channel EMG dataset with the technology of cIONM,
aiming to recognize EMG discharge patterns and predict postoperative outcomes.

3. Data Description

This dataset was collected from West China Hospital of Sichuan University, Chengdu, Sichuan,
China. Cerebellopontine Angle Tumor Surgery (CPA) is one of the most common types of neu-
rosurgical procedures and accounts for approximately 10% of all intracranial tumors [7]. To cap-
ture typical discharge patterns, we chose CPA as a representative type of skull base surgery. We
collected intraoperative EMG data from 11 patients. These signals were produced during their
surgeries. All the patients underwent the same type of surgery.
Compared to previous studies focusing on single-channel EMG signals, we gathered six-
channel EMG data simultaneously. We continuously monitored EMG produced by five cranial
nerves during the whole surgery. In the six channels monitored, channels 1 and 2 monitored
continuous EMG signals from the VII cranial nerve (i.e., the facial nerve). Channel 3 to channel 6
monitored continuous EMG from the V, XI, X, and XII cranial nerves respectively. Table 1 shows
the baseline information of these 11 patients. The detailed structure of this dataset is shown in
Fig. 1, where “readme.pdf” describes in detail how to understand the label and how each seg-
ment of EMG data should correspond to the label. The general folder for the dataset contains
two subfolders, “raw” and “process”. In the “raw” folder, the label folder keeps the label files la-
belled by the neurophysiologists, in which the meanings represented by each label and the EMG
signal waveform are shown in Table 2.
The EMG data of patients 1–11 are stored in the “data” folder, respectively. Due to various
unavoidable factors such as dormancy of the monitoring machine during the operation, the data
of the whole operation is divided into a number of segments. In this dataset, the EMG data of
the patients are divided into 2 to 6 segments, respectively. For example, in the EMG data folder
of patient 1, there are four EMG files, namely, 1_1.txt, 1_2.txt, 1_3.txt, 1_4.txt. The data of other
patients are stored in segments according to this method. However, this is usually not caused
by pathological or meaningful factors, and it does not affect the continuity of the EMG data.
Please note that in order to protect the patient’s privacy, we anonymized the data. Specifically,
 4
Table 2
Classes of EMG signal content.

Event class Activity name Description EMG Signal Waveform

0 A-train A sinusoidal, symmetrical sequence of high-frequency and low-amplitude signals, indicates

W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250

permanent injury

1 A-train salvo A typical pattern of A-trains, as salvos of very short trains

2 B-train with spike Single spikes that may be regularly or irregular occurring

3 B-train with burst Single bursts that may be regularly or irregular occurring

4 Evoke Electric evoked compound muscle action potential of muscles innervated by the cranial nerve

5 Artifact Interference signals caused by grinding drills, static from equipment, etc.

6 Quiet The usual healthy EMG baseline activity

 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250 5

Fig. 1. The record structure of the dataset.

we removed sensitive information such as patient’s name, ID, etc. from the files and replaced
the patientʼs name with a serial number.
In each individual EMG data file, such as “1_1.txt”, the first line contains certain configu-
ration values for the segment of electromyographic signal. “TimePoints” indicates the number
of sampling points in the file. “Channels=6” signifies that we collected data from six channels.
“SamplingInterval[ms]=1.0 0 0” denotes a sampling interval of 1.0 0 0ms, which corresponds to a
sampling frequency of 1 KHz. "Time" refers to the start time. The sampling points and the sam-
pling interval can be used to calculate the duration of the continuous EMG signal collected in the
file. Apart from the first line, the remaining data consists of TimePoints rows with 6 columns,
where each column represents the electromyographic signal collected from a specific channel.
Specifically, both channel 1 and channel 2 capture the EMG signals of the facial nerve, i.e., the
VII cranial nerve. Meanwhile, Channel 3 to channel 6 capture the electromyographic signals of
the V, XI, X, and XII cranial nerves, respectively.
Next, we’ll give an example of how to temporally map a data file to its corresponding la-
bel file. In the first line of “1_1.txt”, we can find that the onset of this segment of EMG signal
as “Time = 10:31:28”. We can also know the number of sampling point is 4,447,200 and the
sampling rate is 10 0 0, from which we can calculate that the signal lasted for 4447.2 s, which
means that the EMG is collected from 10:31:28 to 11:45:07. After the calculation we can look
for the events that occurred during the corresponding time period from the corresponding la-
bel file (1_label.pdf). For example, from 11:24:43 to 11:25:27, the events of the six channels are
shown as 6,6,6,6,2,6.
In addition, to make it easier for researchers to understand and apply this dataset, we also
provided csv files which are more commonly used in deep learning fields in the “process” folder.
Each csv file contains all information of one patient, including six channels of EMG data and
corresponding labels. Researchers are usually more familiar with csv files and can easily utilize
this dataset for their studies.

3.1. Label Description

In this dataset, in order to facilitate subsequent research such as training neural networks
with continuous and complex electromyographic signals collected during surgery for classifi-
cation or prediction tasks, we proposed an innovative data annotation method after exten-
sive discussions and explorations with clinical physicians. As shown in Table 2, we classi-
fied electromyographic signals into seven categories according to previous studies [5,6], where
6 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250

Event “6” represents the usual healthy EMG baseline activity which we defined as "quiet".
We focused more on the six signal categories with special clinical significances, labelled as
Event “0” to “5”. The EMG signal waveform of Event 3 comes from patient 2, and the others are
come from patient 1 in this dataset [8]. The descriptions of the events are derived from discus-
sions with neurophysiologists and previous research [6,9,10]. Additionally, to simplify the record-
ing of event start and end times, we established markers for the start and end of each event,
with 0 indicating the start and 1 indicating the end. The signal category of each channel is rep-
resented by a corresponding number in Table 2. Each label is a string of seven digits. Thus, every
label generated by the physicianʼs annotation follows the format: start/end, channel1, channel2,
..., channel6. During the surgical procedure, if meaningful EMG signals occur, neurophysiologists
mark the event start and end in the MEE-20 0 0 monitoring software before and after the event.
For example, a corresponding label showing “0, 6, 6, 6, 6, 6, 2”means that the EMG signal in
channel 6 shows “B-train with spike” while other channels are “quiet”. Through this innovative
labelling method, we obtained continuous and authentic electromyographic data during surgery,
which assisted the organization of the data.

4. Experimental Design, Materials and Methods

4.1. Patient population

We monitored a consecutive series of 11 patients undergoing Cerebellopontine Angle Tumor

(CPA) Surgery. All of the included patients underwent an acoustic neuroma surgery, which is a
type of the CPA surgery. The mean age of the patients was 49.73 years (range 26–78 years).

4.2. Data collection

Unlike conventional myoelectric acquisition methods [11] used for gesture recognition or
physiological signal acquisition, the intraoperative EMG acquisition process in cIONM is more
continuous and specific. Surgery was performed through a standard retrosigmoid approach. Total
intravenous anaesthesia was administered using propofol and opioids. The cranial nerve (CN) V,
VII, X, XI, XII was monitored with Neuromaster G1 MEE-20 0 0 (Nihon Kohden, Inc., Tokyo, Japan)
as shown in Fig. 2, using free-running EMG, TcMEP during the whole surgery. Paired subdermal
needle electrodes (1500/13/0.4mm) were inserted in the orbicularis oculi, orbicularis oris, mas-
seter, upper trapezius, cricothyroid, genioglossus muscle. The specific positions of the electrodes
are shown in Fig. 3. The numbers in the figure represents the number of channels, and the po-
sitions of the channels also show the positions of electrodes. The Roman numerals in the legend
represent which cranial nerve the channel monitors, and the legend also indicates the muscle to
which the electrode corresponds. The filter was set between 5 and 2 kHz. All data were acquired
using a standard methodology consistent with latest international guidelines. [12–14] Study data
were collected and managed using REDCap electronic data capture tools hosted at West China
Hospital, Sichuan University- Dept. Neurosurgery. [15,16] REDCap (Research Electronic Data Cap-
ture) is a secure, web-based software platform designed to support data capture for research
studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking
data manipulation and export procedures; 3) automated export procedures for seamless data
downloads to common statistical packages; and 4) procedures for data integration and interop-
erability with external sources.

4.3. Data export

The software we used for data visualization and data export is Neuromaster Review Mode
V05-11 (Nihon Kohden, Inc., Tokyo, Japan). To export data of a selected patient, we clicked on
 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250 7

Fig. 2. The device and interface for intraoperative EMG collection.

8 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250

Fig. 3. The positions of subdermal needle electrodes.

Fig. 4. The software interface when exporting data.

the “review” button and then entered the software page. By selecting the "free run" option un-
der the “exam” tab, the “Free run waveforms (0.1s/div)” window appeared, as shown in Fig. 4.
This window displays continuous EMG signals collected from the first to the sixth channels, with
the date and time of surgery displayed in the upper left corner. The time axis window below the
EMG waveform window provides an intuitive way to observe whether continuous EMG signals
are being collected during this time period. Continuous EMG signals are presented as horizon-
tally continuous blue lines in this window.
To export the continuous EMG data, we right-clicked on the “Free run waveforms (0.1s/div)”
window and selected the “Export data to ASCII/BINARY file” option. For convenient data pro-
cessing after export, in the options tab that appeared, we selected the “ASCII” option for the
 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250 9

format and the “Continuous” option for the mode, and kept the range as the default “Current
waveforms” option. Finally, we clicked on the “OK” button to export the selected phase of six-
channel continuous EMG signals. It should be noted that due to various reasons mentioned in
the “DATA DESCRIPTOR” part, it’s necessary to ensure that each segment of the EMG data for
the entire surgery has been exported without omission.
After exporting the EMG data and event list of every participant, we preprocessed the data
and generated csv files containing the labels and EMG data. It should be noted that to protect
the patients’ privacy, the uploaded event lists have been manually processed to remove sensitive
information such as names and IDs.

Limitations

Not applicable

Ethics Statement

All subjects gave written informed consent in accordance with the Declaration of Helsinki.
The study was approved by the Ethics Committee on Biomedical Research, West China Hospital
of Sichuan University, with the Ethics Code: 2020 132. The decision to anonymize the data was
made as part of the ethics approval process.

Credit Author Statement

Wanting Ma: Conceptualization, Methodology, Software, Data curation, Writing - original

draft, Writing - review and editing. Lin Chen: Data curation, Writing - review and editing. Xiao-
fan Pang: Data collection, Data curation, Expert Annotation. Yuanwen Zou: Supervision, Writing
- review and editing.

Data Availability

A Multichannel Continuous Clinical Electromyography Dataset from Neurosurgery (Original

data) (Mendeley Data).

Acknowledgments

This research did not receive any specific grant from funding agencies in the public, commer-
cial, or not-for-profit sectors.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal rela-
tionships that could have appeared to influence the work reported in this paper.

References
[1] C. Topsakal, O. Al-Mefty, K.R. Bulsara, V.S. Williford, Intraoperative monitoring of lower cranial nerves in skull
base surgery: technical report and review of 123 monitored cases, Neurosurg. Rev. 31 (2008) 45–53, doi:10.1007/
s10143- 007- 0105- 5.
10 W. Ma, L. Chen and X. Pang et al. / Data in Brief 53 (2024) 110250

[2] M. Sutter, A. Eggspuehler, A. Muller, J. Dvorak, Multimodal intraoperative monitoring: an overview and proposal of
methodology based on 1017 cases, Eur. Spine J. 16 (2007) 153–161, doi:10.10 07/s0 0586-0 07- 0417- 8.
[3] P. Stankovic, J. Wittlinger, R. Georgiew, N. Dominas, S. Hoch, T. Wilhelm, Continuous intraoperative neuromonitoring
(cIONM) in head and neck surgery—a review, HNO 68 (2020) 86–92, doi:10.10 07/s0 0106- 020- 00824- 1.
[4] J. Romstöck, C. Strauss, R. Fahlbusch, Continuous electromyography monitoring of motor cranial nerves during cere-
bellopontine angle surgery, J. Neurosurg. 93 (20 0 0) 586–593, doi:10.3171/jns.20 0 0.93.4.0586.
[5] X. Zha, L. Wehbe, R.J. Sclabassi, Z. Mace, Y.V. Liang, A. Yu, J. Leonardo, B.C. Cheng, T.A. Hillman, D.A. Chen, C.N. Riv-
iere, A deep learning model for automated classification of intraoperative continuous EMG, IEEE Trans. Med. Robot.
Bionics 3 (2021) 44–52, doi:10.1109/TMRB.2020.3048255.
[6] J. Prell, S. Skinner, EMG monitoring, in: Handbook of Clinical Neurology, Elsevier, 2022, pp. 67–81, doi:10.1016/
B978- 0- 12- 819826- 1.0 0 0 02-8.
[7] N. Canbaz, E. Atılgan, E. Tarakcı, M.G. Papaker, Evaluation of balance after surgery for cerebellopontine angle tumor,
J. Back Musculoskel. Rehab. 32 (2019) 93–99, doi:10.3233/BMR-181198.
[8] W. Ma, C. Lin, X. Pang, A Multichannel Continuous Clinical Electromyography Dataset from Neurosurgery, Mendeley
Data V2, (2024), doi:10.17632/7hyptcbkkd.2.
[9] J. Prell, C. Scheller, S. Simmermacher, C. Strauss, S. Rampp, Facial nerve EMG: low-tech monitoring with a stopwatch,
J. Neurol. Surg. A Cent. Eur. Neurosurg. 82 (2021) 308–316, doi:10.1055/s- 0040- 1701616.
[10] J. Prell, S. Rampp, J. Rachinger, C. Scheller, R. Naraghi, C. Strauss, Spontaneous electromyographic activity during
microvascular decompression in trigeminal neuralgia, J. Clin. Neurophysiol. 25 (2008) 225–232, doi:10.1097/WNP.
0b013e31817f368f.
[11] V. Khodadadi, F.N. Rahatabad, A. Sheikhani, N.J. Dabanloo, A dataset of a stimulated biceps muscle of electromyo-
gram signal by using rossler chaotic equation, Data Br. 49 (2023) 109438. doi:10.1016/j.dib.2023.109438.
[12] W. Cw, C. Fy, D. H, B. Ja, C. Cr, C. Ay, D. Gr, D. Qy, G. Pe, H. Nw, R. Jc, S. Jj, S. Cf, S. Bc, T. Ns, S. Sv, S. Sk, U. Ml, W. I,
W. Rj, R. Gw, International neuromonitoring study group guidelines 2018: Part II: Optimal recurrent laryngeal nerve
management for invasive thyroid cancer-incorporation of surgical, laryngeal, and neural electrophysiologic data, The
Laryngoscope 128 (Suppl 3) (2018), doi:10.1002/lary.27360.
[13] Korean Society of Intraoperative Neurophysiological Monitoring, Korean Neurological Association, Korean academy
of rehabilitation medicine, korean society of clinical neurophysiology, Korean association of EMG electrodiagnos-
tic medicine, clinical practice guidelines for intraoperative neurophysiological monitoring: 2020 update, Ann. Clin.
Neurophysiol. 23 (2021) 35–45.
[14] R. Gw, D. H, A. H, B. M, B. R, B. M, C. B, C. S, C. Fy, D. G, F. C, H. D, K. D, L. K, M. P, M. R, M. A, O. L, P. N, P.
Md, R. A, S. J, S.-S. A, S. T, V.S. S, S. S, T. H, V. E, W. G„ Electrophysiologic recurrent laryngeal nerve monitoring
during thyroid and parathyroid surgery: international standards guideline statement, The Laryngoscope 121 (Suppl
1) (2011), doi:10.1002/lary.21119.
[15] P.A. Harris, R. Taylor, R. Thielke, J. Payne, N. Gonzalez, J.G. Conde, Research electronic data capture (REDCap)—
a metadata-driven methodology and workflow process for providing translational research informatics support, J.
Biomedi. Inform. 42 (2009) 377–381, doi:10.1016/j.jbi.2008.08.010.
[16] P.A. Harris, R. Taylor, B.L. Minor, V. Elliott, M. Fernandez, L. O’Neal, L. McLeod, G. Delacqua, F. Delacqua, J. Kirby,
S.N. Duda, The REDCap consortium: Building an international community of software platform partners, J. Biomed.
Inform. 95 (2019) 103208, doi:10.1016/j.jbi.2019.103208.