The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

actisomni

actisomni estimates sleep from wrist-worn accelerometry. It provides sleep-period-time (SPT) guiders, sustained-inactivity-bout (SIB) labels, consensus sleep labels, diary helpers, and wrappers for the asleep, sleeper, and rester (AcceleRest) machine-learning models.

The guider functions identify a likely main sleep window. Combine one or more guiders with a SIB label when an epoch-level sleep/wake label is needed.

Why not actisleep?

The original package name was actisleep, but that conflicted with a previous R package: https://CRAN.R-project.org/package=ActiSleep.

Installation

Install the development version from GitHub:

# install.packages("remotes")
remotes::install_github("jhuwit/actisomni")

Sleep guiders and labels

For epoch-level activity data, use an activity-based guider such as L5 and then apply a SIB label. The result has one logical sleep label per epoch.

library(actisomni)

time <- as.POSIXct("2020-01-01 18:00:00", tz = "UTC") + 0:1439 * 60
epochs <- data.frame(
  time = time,
  activity = c(rep(20, 240), rep(0, 420), rep(20, 780)),
  sib = c(rep(FALSE, 240), rep(TRUE, 420), rep(FALSE, 780))
)

guider <- acti_sleep_l5(epochs)
epochs$sleep <- acti_sleep_label(epochs$sib, guider)
head(epochs)
#>                  time activity   sib sleep
#> 1 2020-01-01 18:00:00       20 FALSE FALSE
#> 2 2020-01-01 18:01:00       20 FALSE FALSE
#> 3 2020-01-01 18:02:00       20 FALSE FALSE
#> 4 2020-01-01 18:03:00       20 FALSE FALSE
#> 5 2020-01-01 18:04:00       20 FALSE FALSE
#> 6 2020-01-01 18:05:00       20 FALSE FALSE

For raw X/Y/Z acceleration, acti_sleep_sib() calculates the van Hees SIB indicator and acti_sleep_ensemble() combines compatible guiders into a consensus label. The example recording supplied by actiread can be used directly:

library(actiread)

raw <- actiread::acti_read_gt3x(
  actiread::acti_example_gt3x(),
  verbose = FALSE
)
result <- acti_sleep_ensemble(raw, epoch = "5 seconds")

result$label_data

result$labels retains one label per guider, while result$fused contains the consensus probability and logical consensus label.

Diary windows

When diary times are available, create a guider directly from onset and wake times. Times are inclusive at both endpoints.

diary_window <- acti_sleep_diary(
  epochs,
  onset = as.POSIXct("2020-01-01 22:00:00", tz = "UTC"),
  wakeup = as.POSIXct("2020-01-02 06:00:00", tz = "UTC")
)
sum(diary_window$window)
#> [1] 481

Machine-learning model wrappers

acti_asleep(), acti_sleeper(), and acti_rester() require complete, strictly time-ordered raw triaxial data. They return the same columns: time, sleep, sleep_probability, sleep_stage, nonwear, and method.

The models and their Python requirements are managed by the upstream packages. For sleeper, download the model files first and provide their directory as model_dir. For rester, run rester::ar_setup() once to download the AcceleRest code and weights; acti_rester() also returns sleep stages and needs at least about 2.1 hours of data.

asleep_estimate <- acti_asleep(raw, verbose = FALSE)
sleeper_estimate <- acti_sleeper(raw, model_dir = "path/to/sleeper-models")
rester_estimate <- acti_rester(raw)

Funding

This work was supported by NIH P30AG021334 and the Johns Hopkins Older Americans Independence Center (OAIC) Pepper Center grant.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.