kielmat 0.0.5

The KielMotionAnalysisToolbox (KielMAT) is a Python-based toolbox for processing motion data.

Welcome to the KielMotionAnalysisToolbox (KielMAT). We are a Python based toolbox for processing motion data.

The toolbox is aimed at motion researchers who want to use Python-based open-source software to process their data. We have implemented validated algorithms in modules to process motion data, as shown in the table below:

Module	Description	Data
Gait sequence detection (GSD)	Detects gaits	3D accelerations from the lower back
Initial contact detection (ICD)	Detects initial contact during gait	3D accelerations from the lower back
Postural transition detection (SSD)	Detects sit-to-stand and stand-to-sit movements	3D accelerations and gyroscope from the lower back
Physical activity monitoring (PAM)	Monitors physical activity levels	3D accelerations from the wrist
More to follow...	Additional modules to be added

The idea is that various motion data can be loaded into our dedicated dataclass which rely on principles from the Motion-BIDS standard.

Data classes

Data classes: conceptual framework

Motion data is recorded with many different systems and modalities, each with their own proprietary data format. KielMAT deals with this by organizing both data and metadata in a BIDS-like format. The BIDS format suggests that motion recording data from a single tracking system is organized in a single *_tracksys-<label>_motion.tsv file.

!!! note

A tracking system is defined as a group of motion channels that share hardware properties (the recording device) and software properties (the recording duration and number of samples).

In KielMAT, data from a single tracking system is therefore loaded into a single pandas.DataFrame. The column headers of this pandas.DataFrame refer to the channels, and the corresponding channels information is likewise available as a pandas.DataFrame.

Similarly, if any events are available for the given recording, these are loaded into a single pandas.DataFrame for each tracking system as well. The events derived from the toolbox can be exported to a BIDS like '*_events.tsv' file.

Data classes: in practice

These concepts are translated into a KielMAT dataclass for each recording: KielMATRecording:

classDiagram
   class KielMATRecording {
      data: dict[str, pd.DataFrame]
      channels: dict[str, pd.DataFrame]
      info: None | dict[str, Any] = None
      events: None | dict[str, pd.DataFrame] = None
      events_info: None | dict[str, Any] = None
      add_events(tracking_system, new_events)
      add_info(key, value)
      export_events(file_path, tracking_system=None, file_name=None, bids_compatible_fname=False)
   }

A recording consists of the motion data from one or more tracking systems, where each tracking system may consist motion data from one or more tracked points. Therefore, the motion data (KielMATRecording.data) are organized as a dictionary where the dictionary keys refer to the tracking systems, and the corresponding values the actual (raw) data as a pandas.DataFrame. The description of data channels (KielMATRecording.channels) is availabe as a dictionary with the same keys, and the values contain the channels description.

>>> from kielmat.datasets import mobilised
>>> file_name = "/mnt/neurogeriatrics_data/Mobilise-D/rawdata/sub-3011/Free-living/data.mat"
>>> recording = mobilised.load_recording(file_name, tracking_systems=["SU", "SU_INDIP"], tracked_points=["LowerBack"])
>>> recording.data
{'SU':         LowerBack_ACCEL_x  ...  LowerBack_BARO_n/a
0                0.967784  ...         1011.628100
1                0.969667  ...         1011.628400
...                   ...  ...                 ...
993022           0.970579  ...         1012.078703
993023           0.960542  ...         1002.580321

[993024 rows x 10 columns], 

'SU_INDIP':      LowerBack_ACCEL_x  ... LowerBack_MAGN_z
0                0.967986  ...         -5.902833
1                0.963671  ...          9.501037
...                   ...  ...               ...
993022           0.951656  ...        -17.987983
993023           0.955107  ...        -18.050600

[993024 rows x 9 columns]
}

>>> recording.channels
{'SU':                  
   name                 type  component   tracked_point  units    sampling_frequency
0  LowerBack_ACCEL_x    Acc   x           LowerBack      g        100.0
1  LowerBack_ACCEL_y    Acc   y           LowerBack      g        100.0
2  LowerBack_ACCEL_z    Acc   z           LowerBack      g        100.0
3  LowerBack_ANGVEL_x   Gyr   x           LowerBack      deg/s    100.0
4  LowerBack_ANGVEL_y   Gyr   y           LowerBack      deg/s    100.0
5  LowerBack_ANGVEL_z   Gyr   z           LowerBack      deg/s    100.0
6  LowerBack_MAGN_x     Mag   x           LowerBack      µT       100.0
7  LowerBack_MAGN_y     Mag   y           LowerBack      µT       100.0
8  LowerBack_MAGN_z     Mag   z           LowerBack      µT       100.0
9  LowerBack_BARO_n/a   Bar   n/a         LowerBack      hPa      100.0, 

'SU_INDIP':
   name                 type  component   tracked_point  units    sampling_frequency
0  LowerBack_ACCEL_x    Acc   x           LowerBack      g        100.0
1  LowerBack_ACCEL_y    Acc   y           LowerBack      g        100.0
2  LowerBack_ACCEL_z    Acc   z           LowerBack      g        100.0
3  LowerBack_ANGVEL_x   Gyr   x           LowerBack      deg/s    100.0
4  LowerBack_ANGVEL_y   Gyr   y           LowerBack      deg/s    100.0
5  LowerBack_ANGVEL_z   Gyr   z           LowerBack      deg/s    100.0
6  LowerBack_MAGN_x     Mag   x           LowerBack      µT       100.0
7  LowerBack_MAGN_y     Mag   y           LowerBack      µT       100.0
8  LowerBack_MAGN_z     Mag   z           LowerBack      µT       100.0
9  LowerBack_BARO_n/a   Bar   n/a         LowerBack      hPa      100.0,
}

!!! note

In the examples you find a [tutorial (the basics of KielMAT)](https://neurogeriatricskiel.github.io/KielMAT/examples/00_tutorial_basics/) that explains the basics of the dataclass and how to work with them.

Installation

The toolbox has been released on pypi and can be installed via pip:

pip install kielmat

It requires Python 3.10 or higher.

Contributing

We welcome contributions to KielMAT! Please refer to our contributing guide for more details.

Authors

Masoud Abedinifar, Julius Welzel, Walter Maetzler, Clint Hansen & Robbin Romijnders