Note
Go to the end to download the full example code
GSD Paraschiv-Ionescu#
This example shows how to use the two variants of the Gait Sequence Detection (GSD) algorithm by
Paraschiv-Ionescu et al.
The normal version called GsdIonescu uses a fixed signal activity threshold and a simplified filter chain and the
adaptive version called GsdAdaptiveIonescu, uses an “automatic” threshold calculation and a more complex filter
chain.
We start by defining some helpers for plotting and loading the data. You can skip them for now and jump directly to “Performance on a single lab trial”, if you just want to see how to apply the algorithm.
# Plotting Helper
# ---------------
# We define a helper function to plot the results of the algorithm.
# Just ignore this function for now.
import json
import matplotlib.pyplot as plt
import pandas as pd
from mobgap import PACKAGE_ROOT
def plot_gsd_outputs(data, **kwargs):
fig, ax = plt.subplots()
ax.plot(data["acc_x"].to_numpy(), label="acc_x")
color_cycle = iter(plt.rcParams["axes.prop_cycle"])
y_max = 1.1
plot_props = [
{
"data": v,
"label": k,
"alpha": 0.2,
"ymax": (y_max := y_max - 0.1),
"color": next(color_cycle)["color"],
}
for k, v in kwargs.items()
]
for props in plot_props:
for gsd in props.pop("data").itertuples(index=False):
ax.axvspan(
gsd.start, gsd.end, label=props.pop("label", None), **props
)
ax.legend()
return fig, ax
Loading some example data#
Note
More infos about data loading can be found in the data loading example.
We load example data from the lab dataset together with the INDIP reference system. We will use the INDIP “WB” output as ground truth. Note, that the “WB” (Walking Bout) output is further processed than a normal “Gait Sequence”. This means we expect Gait Sequences to contain some false positives compared to the “WB” output. However, a good gait sequence detection algorithm should have high sensitivity (i.e. contain all the “WBs” of the reference system).
We also load the original Matlab results for the adaptive version of the algorithm.
from mobgap.data import LabExampleDataset
lab_example_data = LabExampleDataset(reference_system="INDIP")
def load_matlab_output(datapoint):
p = datapoint.group_label
with (
PACKAGE_ROOT.parent
/ f"example_data/original_results/gsd_adaptive_ionescu/lab/{p.cohort}/{p.participant_id}/GSDB_Output.json"
).open() as f:
original_results = json.load(f)["GSDB_Output"][p.time_measure][p.test][
p.trial
]["SU"]["LowerBack"]["GSD"]
if not isinstance(original_results, list):
original_results = [original_results]
return (
(
pd.DataFrame.from_records(original_results).rename(
{"Start": "start", "End": "end"}, axis=1
)[["start", "end"]]
* datapoint.sampling_rate_hz
)
.round()
.astype("int64")
)
Performance on a single lab trial#
Below we apply the algorithm to a lab trail, where we only expect a single gait sequence.
For that we load the relevant data pieces.
Note, that we use to_body_frame to convert the data to body frame coordinates.
This is possible, as we know the data is well aligned with the defined sensor frame convention.
However, technically, this step (or any alignment step) is not required for the GsdIonescu algorithm variants, as
they both work on the Acc norm and can hence be used without prior alignment.
Hence, the algorithms support passing data with either sensor or body frame column naming.
from mobgap.gait_sequences import GsdAdaptiveIonescu, GsdIonescu
short_trial = lab_example_data.get_subset(
cohort="MS", participant_id="001", test="Test5", trial="Trial2"
)
short_trial_matlab_output = load_matlab_output(short_trial)
short_trial_reference_parameters = short_trial.reference_parameters_.wb_list
short_trial_output_normal = GsdIonescu().detect(
short_trial.data_ss, sampling_rate_hz=short_trial.sampling_rate_hz
)
short_trial_output_adaptive = GsdAdaptiveIonescu().detect(
short_trial.data_ss, sampling_rate_hz=short_trial.sampling_rate_hz
)
print("Reference Parameters:\n\n", short_trial_reference_parameters)
print("\nMatlab Adaptive Ionescu Output:\n\n", short_trial_matlab_output)
print("\nPython Normal Ionescu Output:\n\n", short_trial_output_normal.gs_list_)
print(
"\nPython Adaptive Ionescu Output:\n\n",
short_trial_output_adaptive.gs_list_,
)
Reference Parameters:
start end ... avg_stride_length_m termination_reason
wb_id ...
0 434 874 ... 1.10562 Pause
[1 rows x 9 columns]
Matlab Adaptive Ionescu Output:
start end
0 438 1115
Python Normal Ionescu Output:
start end
gs_id
0 385 972
Python Adaptive Ionescu Output:
start end
gs_id
0 442 885
When we plot the output, we can see that the python version is more accurate and cuts the gait sequence roughly at the same time as the reference system, while the matlab version calssifies the small movement after the gait sequence as a gait as well.
fig, ax = plot_gsd_outputs(
short_trial.data_ss,
reference=short_trial_reference_parameters,
matlab_adaptive=short_trial_matlab_output,
python_adaptive=short_trial_output_adaptive.gs_list_,
python_normal=short_trial_output_normal.gs_list_,
)
fig.show()
Performance on a longer lab trial#
Below we apply the algorithm to a lab trail that contains activities of daily living. This is a more challenging scenario, as we expect multiple gait sequences.
long_trial = lab_example_data.get_subset(
cohort="MS", participant_id="001", test="Test11", trial="Trial1"
)
long_trial_matlab_output = load_matlab_output(long_trial)
long_trial_reference_parameters = long_trial.reference_parameters_.wb_list
long_trial_output_normal = GsdIonescu().detect(
long_trial.data_ss, sampling_rate_hz=long_trial.sampling_rate_hz
)
long_trial_output_adaptive = GsdAdaptiveIonescu().detect(
long_trial.data_ss, sampling_rate_hz=long_trial.sampling_rate_hz
)
print("Reference Parameters:\n\n", long_trial_reference_parameters)
print("\nMatlab Adaptive Ionescu Output:\n\n", long_trial_matlab_output)
print("\nPython Normal Ionescu Output:\n\n", long_trial_output_normal.gs_list_)
print(
"\nPython Adaptive Ionescu Output:\n\n", long_trial_output_adaptive.gs_list_
)
/home/docs/checkouts/readthedocs.org/user_builds/mobgap/checkouts/v0.9.0/mobgap/data/_mobilised_matlab_loader.py:1082: UserWarning: There were multiple ICs with the same index value, but different LR labels. This is likely an issue with the reference system you should further investigate. For now, we set the `lr_label` of the stride corresponding to this IC to Nan. However, both values still remain in the IC list.
return parse_reference_parameters(
Reference Parameters:
start end ... avg_stride_length_m termination_reason
wb_id ...
0 1019 1768 ... 0.942678 Pause
1 4534 5549 ... 0.483923 Pause
2 9665 10569 ... 0.506458 Pause
3 12337 14633 ... 0.803933 Pause
4 20151 20982 ... 0.507484 Pause
5 21378 22129 ... 0.599360 Pause
[6 rows x 9 columns]
Matlab Adaptive Ionescu Output:
start end
0 807 1842
1 5205 6010
2 9545 10620
3 12988 14670
4 20085 22728
Python Normal Ionescu Output:
start end
gs_id
0 847 1940
1 5202 6052
2 9585 10675
3 12952 14725
4 20235 22475
Python Adaptive Ionescu Output:
start end
gs_id
0 802 2032
1 4485 6092
2 9552 10730
3 11402 12040
4 12882 14752
5 19952 22507
When we plot the output, we can see that the python version is more sensitive. It detects longer gait sequences and even one entire gait sequence that is not detected by the matlab version. But, like before, the Python version seems to provide the better results when compared to the reference system.
fig, _ = plot_gsd_outputs(
long_trial.data_ss,
reference=long_trial_reference_parameters,
matlab=long_trial_matlab_output,
python_normal=long_trial_output_normal.gs_list_,
python_adaptive=long_trial_output_adaptive.gs_list_,
)
fig.show()
Evaluation of the algorithm against a reference#
To quantify how the Python output compares to the reference labels, we are providing a range of evaluation functions. See the example on GSD evaluation for more details.
Total running time of the script: (0 minutes 1.720 seconds)
Estimated memory usage: 9 MB