"Preprocess data with spike2py."
Project description
spike2py_preprocess provides a simple way to batch (pre)process data with spike2py.
spike2py_preprocess can be used to batch read a series of .mat
files and save them to .pkl
files.
However, the power of spike2py_preprocess is its ability to also preprocess the data, and this for a single trial, all trials from a subject, or all trials from a study.
Moreover, spike2py_preprocess can be used to extract only relevant sections of data; simply add two Spike2 TextMarks to mark the section of data to be extracted.
More than one section can be extracted per trial.
Trial
In python:
>>> from spike2py.trial import TrialInfo
>>> from spike2py_preprocess.trial import trial
>>> trial_info = TrialInfo(file="0004.mat",
name='h_reflex_curve',
subject_id='sub01',
path_save_trial='./proc')
>>> trial(trial_info)
On the command line:
$ python -m spike2py_preprocess trial --help
$ python -m spike2py_preprocess trial <path_to_trial_info_json>
or simply:
$ spike2py_preprocess trial --help
$ spike2py_preprocess trial <path_to_trial_info_json>
Here, we need to point spike2py_preprocess.py
to a valid json file.
The json file requires the following fields:
{
"file": "/home/maple/study/sub01/data/raw/sub01_DATA000_H_B.mat",
"channels": ["FDI", "W_EXT", "stim"],
"name": "biphasic_high_fq",
"subject_id": "sub01",
"path_save_trial": "/home/maple/study/sub01/data/proc"
}
Subject:
In Python:
>>> from spike2py_preprocess.subject import subject
>>> from pathlib import Path
>>> subject_folder = Path('sub01')
>>> subject(subject_folder)
On the command line:
$ python -m spike2py_preprocess subject --help
$ python -m spike2py_preprocess subject /home/maple/study/sub01
or simply:
$ spike2py_preprocess subject --help
$ spike2py_preprocess subject /home/maple/study/sub01
Study:
In Python:
>>> from spike2py_preprocess.study import study
>>> from pathlib import Path
>>> study_folder = Path('great_study')
>>> study(study_folder)
On the command line:
$ python -m spike2py_preprocess study --help
$ python -m spike2py_preprocess study /home/maple/study/
or simply:
$ spike2py_preprocess study --help
$ spike2py_preprocess study /home/maple/study/
Preprocess
You can specify the preprocessing settings to apply to one or more channels by including one or more <level>_preprocess.json
files.
For a single trial, spike2py_preprocess looks for <trialname.mat>_preprocess.json
in the same folder as the .mat
file.
For all trials for a subject, spike2py_preprocess looks for subject_preprocess.json
in the provided subject folder.
Finally, for all trials in a study, spike2py_preprocess looks for study_preprocess.json
in the provided study folder.
Controlling the preprocessing
By including study_preprocess.json
, subject_preprocess.json
and <trialname.mat>_preprocess.json
files in a given file structure,
it is possible to provide a general preprocess scheme, but that can be overridden for a given subject or a given trial.
Below is an example of what could be included in a preprocess file.
As you can see, each item is the name of a channel that exists in the .mat
file.
For each channel, a specific preprocessing step is specified.
At present, the preprocessing steps that are possible are those included in the spike2py
SignalProcessing
mixin class.
The keys are the call to the preprocessing method, and the values are the inputs to those methods (if required).
{
"Fdi": {
"remove_mean": "",
"lowpass": "cutoff=200"
},
"W_Ext": {
"remove_mean": "",
"bandstop": "cutoff = [49, 51]"
},
"Stim": {
"lowpass": "cutoff=20, order=8"
}
}
IMPORTANT
Note that, in our example above, the channels were specified as "channels": ["FDI", "W_EXT", "stim"]
,
but here they are specified as Fdi
, W_Ext
, and Stim
. The reason for this is the researchers often used a wide
variety of styles to label channels, sometimes with ALLCAPS, other times with camel_case, or a combination of many.
To standardise things, spike2py
applied TitleCase to each of the channel names. If you are not 100% certain what the
resulting channel name will be, simply apply spike2py_preprocess
with you preprocess specified. This will result in a
pickle (.pkl
) file that can be opened with spike2py
.
import spike2py as s2p
from pathlib import Path
tutorial = s2p.trial.load(file=Path('data.pkl'))
Now if you simply type tutorial
and hit return, you will see information about the data, include the channel names.
File structure
Below is an example of the required file/folder structure for spike2py_preprocess.
In the example, sub02_DATA000_H_B.mat
has its own preprocess details located in preprocess_sub02_DATA000_H_B.json.
Similarly, at the subject level, sub02
has a subject_preprocess.json
file. This means all their files (excluding sub02_DATA000_H_B.mat
) will be preprocessed in the same way.
Finally, because sub01
does not include a dedicated .json
file, their data would simply be read and saved as .pkl
files if their data was analysed on their own.
However, if spike2py_preprocess was used to preprocess all trials in the study, trials from sub01
would be preprocessed with the details provided in study_preprocess.json
.
Study folder structure and required files
Below is an example study file structure. The study folder can have whatever name you like, study1
in this case.
Similarly, the subject folders can have whatever names you like, but they should match the list of subjects you include
in the study_info.json
file (see below for more details).
The raw data, in .mat
format for each subject must be located in a folder with the name raw
.
study1/
├── study_info.json
├── study_preprocess.json
├── sub01
│ ├── raw
│ │ ├── sub01_DATA000_H_B_trial_info.json
│ │ ├── sub01_DATA000_H_B.mat
│ │ ├── sub01_DATA001_C_B.mat
│ │ ├── sub01_DATA002_C_M.mat
│ │ └── sub01_DATA003_H_M.mat
│ └── subject_info.json
└── sub02
├── raw
│ ├── preprocess_sub02_DATA000_H_B.json
│ ├── sub02_DATA000_H_B.mat
│ ├── sub02_DATA001_C_B.mat
│ ├── sub02_DATA002_C_M.mat
│ └── sub02_DATA003_H_M.mat
├── subject_info.json
└── subject_preprocess.json
subject_info.json
This file contains details about the subject. Additional information can appear in this file, but at a minimum it requires that "subject_id" be provided, as well as "trials", which contains the various trials to be processed for this subject. For each trial, the minimum data required is "name" and "file". If "channels" is provided, only these channels will be included and preprocessed; if not provided, all channels will be included.
{
"subject_id": "sub01",
"age": 50,
"gender": "F",
"trials": {
"trial1": {
"name": "conv_biphasic",
"file": "sub01_001.mat"
},
"trial2": {
"name": "khz_biphasic",
"file": "sub01_002.mat",
"channels": ["FDI", "W_EXT", "stim"]
}
}
}
study_info.json
This file contains details about the study. Additional information can appear in this file, but at a minimum it requires that "name" and "subjects" be provided. If "channels" is provided, only these channels will be included and preprocessed, noting that this can be trumped
{
"name": "TSS_H-reflex",
"subjects": [
"sub01",
"sub02"
],
"channels": ["FDI", "W_EXT", "stim"]
}
Spike2 TextMarks
Please refer to the document entitled "How_to_add_TextMarks_in_Spike2.pdf" for a guide on how to add TextMarks in Spike2.
If you add two TextMarks with the same label (e.g. 'MVC'), the section of data between the two TextMarks will be extracted and saved to a .pkl file. Many such pairs of TextMarks can be included in a trial.
If you have two related sections of data, but want to exclude a middle section that is not useful or relevant, you can add four labels, two around each section of data of interest, that have the same label, the data from both sections will be concatenated and extracted.
Note that Spike2 TextMarks need to be added prior to batch exporting the trial to .mat.
Installing
spike2py_preprocess is available on PyPI:
$ python -m pip install spike2py_preprocess
spike2py officially supports Python 3.8+.
Contributing
Like this project? Want to help? We would love to have your contribution! Please see CONTRIBUTING to get started.
Code of conduct
This project adheres to the Contributor Covenant code of conduct. By participating, you are expected to uphold this code. Please report unacceptable behavior to heroux.martin@gmail.com.
License
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for spike2py_preprocess-0.1.14.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | ce3667367553b2ed602ca9a82070b7ef5906a99916cd87e47e82720baa62b581 |
|
MD5 | 1389381048bdb57320e4838cf385de8c |
|
BLAKE2b-256 | d31d9fbed1ef08b6e4f4a1d0744e09eea89783a68c4dd562c613e0903d23ef0c |
Hashes for spike2py_preprocess-0.1.14-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8d162e89e0b8b858e42f1019d2d15d17a17da830449c7c3225b3785da1b827c7 |
|
MD5 | db090f0afb67ab78d17051d57b4a01ff |
|
BLAKE2b-256 | aab1339367a36eba8d095319e573530e6b1f7ad46e4aeeb0d26395c29575695c |