A python package for extracting actigraphy counts from accelerometer data.
pip install agcounts
Download test data:
curl -L https://github.com/actigraph/agcounts/files/8247896/GT3XPLUS-AccelerationCalibrated-1x8x0.NEO1G75911139.2000-01-06-13-00-00-000-P0000.sensor.csv.gz --output data.csv.gz
Run a simple test
import pandas as pd
import numpy as np
from agcounts.extract import get_counts
def get_counts_csv(
file,
freq: int,
epoch: int,
fast: bool = True,
verbose: bool = False,
time_column: str = None,
):
if verbose:
print("Reading in CSV", flush=True)
raw = pd.read_csv(file, skiprows=0)
if time_column is not None:
ts = raw[time_column]
ts = pd.to_datetime(ts)
time_freq = str(epoch) + "S"
ts = ts.dt.round(time_freq)
ts = ts.unique()
ts = pd.DataFrame(ts, columns=[time_column])
raw = raw[["X", "Y", "Z"]]
if verbose:
print("Converting to array", flush=True)
raw = np.array(raw)
if verbose:
print("Getting Counts", flush=True)
counts = get_counts(raw, freq=freq, epoch=epoch, fast=fast, verbose=verbose)
del raw
counts = pd.DataFrame(counts, columns=["Axis1", "Axis2", "Axis3"])
counts["AC"] = (
counts["Axis1"] ** 2 + counts["Axis2"] ** 2 + counts["Axis3"] ** 2
) ** 0.5
ts = ts[0 : counts.shape[0]]
if time_column is not None:
counts = pd.concat([ts, counts], axis=1)
return counts
def convert_counts_csv(
file,
outfile,
freq: int,
epoch: int,
fast: bool = True,
verbose: bool = False,
time_column: str = None,
):
counts = get_counts_csv(
file, freq=80, epoch=60, verbose=True, time_column=time_column
)
counts.to_csv(outfile, index=False)
return counts
counts = get_counts_csv("data.csv.gz", freq=80, epoch=60)
counts = convert_counts_csv(
"data.csv.gz",
outfile="counts.csv.gz",
freq=80,
epoch=60,
verbose=True,
time_column="HEADER_TIMESTAMP",
)