Tool for assessing postural stability/balance/sway in Virtual Reality.
The oscillating room condition
Assess posture in one of three conditions.
- Normal (vision).
- No vision (HMD goes dark, ask participant to close their eyes too).
- Oscillating room. Room pitch rotation oscillates at a amplitude and period that you specify.
The task measures the head position using the built-in tracking of the HMD, which has high accuracy (Validation for Vive, and Oculus Rift S).
Built with Unity Experiment Framework.
The virtual room used for the assessments
This software has been used in the Brookes et al. 2019 where we found significant differences between age groups and between the vision and oscillating room condition.
Fig. 7 from Brookes et al. 2019
This is the repository for the Unity project source code; clone/fork and edit as necessary in Unity version 2018.1.X.
If you just want to download and run the assessment, first download the latest Release Build, extract the zip file to your Windows PC, and follow the instructions below:
- Make sure you have a SteamVR compatible headset set up on your PC.
- Put participant in a headset standing in the center of your VR area.
- Launch
posture-assessment-vr.exe - Create a new participant list (or select existing) - data will be stored next to wherever you put this list.
- Enter the participant details.
- Press Start.
- From here you can perform measurements:
- If you want height/armspan measures, ask participant to stand with controllers in hands, hold them out horizontally.
- Click the buttons to record height/armspan. Height and armspan will be added to the participant's row in the participant list.
- Click buttons on-screen to perform the various assessments as needed. Tell participants to look at the fixation cross.
The experimenter's user interface
Info:
This project is built with SteamVR. It should work with any SteamVR compatible headset and SteamVR installed.
You can change some parameters by modifying the file: posture-assessment-vr_Data/StreamingAssets/psat-vr.json.
Data output is just raw movement data. Calculate path length by summing the point-to-point distances observed in head movement.
Here is a commented example of an R script (written in Tidyverse style) that will process the data from this task, calculating the path length (a measure of stability - lower is better).
library(tidyverse)
# our data is stored in a folder called `data`.
# within the `data` folder is our experiment folder (i.e. `psat-vr` by default).
# within that, is each participant folder.
# within those, is the session folder for each participant.
# first, we load all the trial_results files.
# we do this by recursively searching all files that match a pattern.
# then reading all the files we find (with `map`) and binding them row-wise (`_dfr`)
posture_trials <- list.files(
path = "data",
pattern = "trial_results.csv",
full.names = TRUE,
recursive = TRUE
) %>%
map_dfr(read_csv)
# output (in my case there is only one session)
# A tibble: 3 x 11
# directory experiment ppid session_num trial_num block_num trial_num_in_bl~ start_time end_time assessment_type
# <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>
# 1 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 2 psat-vr-~ psat-vr-1~ P_19~ 1 2 1 2 55.4 65.4 NoVision
# 3 psat-vr-~ psat-vr-1~ P_19~ 1 3 1 3 73.3 83.3 RoomOscillate
# ... with 1 more variable: center_eye_movement_filename <chr>
# now we want the head movement data for every trial.
# we can do this by reading each filename given in the `*_movement_filename` column.
# then we can use `unnest` to give 1 row per timestep.
posture_movement <- posture_trials %>%
mutate(center_eye_movement = map(file.path("data", directory, center_eye_movement_filename), read_csv)) %>%
unnest()
# output (notice many more rows, and the `pos_*` columns)
# # A tibble: 2,310 x 18
# directory experiment ppid session_num trial_num block_num trial_num_in_bl~ start_time end_time assessment_type
# <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>
# 1 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 2 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 3 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 4 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 5 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 6 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 7 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 8 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 9 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# 10 psat-vr-~ psat-vr-1~ P_19~ 1 1 1 1 37.0 47.0 Vision
# # ... with 2,300 more rows, and 8 more variables: center_eye_movement_filename <chr>, time <dbl>, pos_x <dbl>,
# # pos_y <dbl>, pos_z <dbl>, rot_x <dbl>, rot_y <dbl>, rot_z <dbl>
# we can now write a function which calculates path length and summarise by it.
# grouping will give us 1 row per trial again
calculate_path_length <- function(x, y, z) {
# calculates sum of point-to-point distances.
sum(
(diff(x) ^ 2 +
diff(y) ^ 2 +
diff(z) ^ 2) ^ 0.5
)
}
posture_summary <- posture_movement %>%
group_by(experiment, ppid, session_num, trial_num, assessment_type) %>%
summarise(path_length = calculate_path_length(pos_x, pos_y, pos_z))
# output (my data are just of 1 participant, 1 session)
# # A tibble: 3 x 6
# # Groups: experiment, ppid, session_num, trial_num [3]
# experiment ppid session_num trial_num assessment_type path_length
# <chr> <chr> <dbl> <dbl> <chr> <dbl>
# 1 psat-vr-10s P_190905151224 1 1 Vision 0.166
# 2 psat-vr-10s P_190905151224 1 2 NoVision 0.118
# 3 psat-vr-10s P_190905151224 1 3 RoomOscillate 0.104
# we can plot these data on a graph
ggplot(posture_summary, aes(x = assessment_type, y = path_length, color = assessment_type)) +
geom_point()
This is just example data, don't pay attention to the pattern.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent