Overview

Read me

About

This is the download site for Nursing Activity dataset:

In the cardiovascular center of a hospital, we collected mobile-sensor data from nurses. We asked the nurses to place mobile devices (iPod Touches) that record accelerations into their breast pockets in a roughly fixed direction. They also attached two small accelerometer devices on their right wrist and the other to the back of their waist. Each sensor measured accelerations on three axes in the range of ±2G.

It includes labeled data for 2 weeks, and unlabelled data for some of the duty days in 2 years.

Download

You can get the files the files tab or here:
http://redmine.sozolab.jp/projects/nursing/files

Citation request

  • [Ubicomp2015] Sozo Inoue, Naonori Ueda, Yasunobu Nohara, Naoki Nakashima, "Mobile Activity Recognition for a Whole Day: Recognizing Real Nursing Activities with Big Dataset",ACM Int'l Conf. Pervasive and Ubiquitous Computing (UbiComp), pp. 1269-1280, 2015-09-09, Osaka.

Files/Directories

  • acts.csv : the list of activity classes.
  • labelled.zip : Sensor data of 3 accelerometers with labels for 2 weeks.
    • Labelled/
      • sensors/#{nurseID}_#{date}_#{start_t}-#{end_t}.csv : sensor data for the nurse of ID #{nurseID} from time #{start_t} (hhmmus) to #{end_t} (hhmmss) on the date #{date} (YYYYMMDD).
      • labels/ :
        • #{nurseID}_#{date}.csv : label data for the nurse of ID #{nurseID} on the date #{date} (YYYYMMDD). Note that a single label file may corresponds to many sensor data file.
  • unlabelled.zip : Unlabelled sensor data of 3 accelerometers without labels for longer term which corresponds to the duty records.
    • Unlabelled/
      • duties.csv : nurses' duty records extracted from the hospital information system.
      • sensors/#{duty_id}.csv: sensor data for the row #{duty_id} of duties.csv.

Formats

Activity classes ( acts.csv )

This is the predefined activity classes.

Example of a CSV file:

"action_id","action_name" 
1,"Anamnese (patient sitting)" 
2,"  Anamnese (nurse standing)" 
3,"Anamnese (nurse half-sitting)" 
4,"  Measure height" 
5,"Measure weight (dorsal)" 
6,"Measure weight (stand)" 
7,"Measure waist" 
8,"Measure blood pressure (dorsal)" 
9,"Measure blood pressure (sit)" 
......

Each column is as follows:

  • action_id: The activity class ID. 41 in total.
  • action_name: The activity class name.

Labelled sensor data

Merged sensor data for iPod on the chest, accelerator on the waist, and accelerator on the wrist.
The 3 types of sensor data were firstly synchronized the time firstly and merged to a one file.

Example of a CSV file:

time,chest_x,chest_y,chest_z,waist_x,waist_y,waist_z,right_x,right_y,right_z
0.114,0.12352,-0.43835,-0.98593,0.09375,0.703125,0.390625,-0.34375,-0.875,-0.21875
0.172,0.06366,-0.41920,-0.84056,-0.078125,1.1875,0.265625,-0.34375,-0.875,-0.21875
0.219,0.06711,-0.41669,-0.89584,0.015625,0.734375,0.15625,-0.28125,-0.9375,-0.28125
0.273,0.08684,-0.40904,-0.91821,-0.171875,0.8125,0.15625,-0.3125,-0.96875,-0.25
0.320,0.06691,-0.40910,-0.87210,-0.109375,0.96875,0.078125,-0.34375,-1.1875,-0.34375
0.371,0.08368,-0.43262,-0.89960,0,0.984375,0.03125,-0.40625,-1.1875,-0.3125
0.420,0.07680,-0.44164,-0.89383,-0.078125,0.78125,-0.09375,-0.46875,-0.875,-0.21875
0.471,0.06476,-0.42001,-0.85538,-0.234375,0.90625,0.109375,-0.53125,-0.78125,-0.0625
0.523,0.09027,-0.43649,-0.87379,-1.01563,1.53125,0.609375,-0.53125,-0.96875,0.03125
......

Each column is as follows:

  • time: relative time [seconds] from the #{start_t} in the file name.
  • chest_x: X axis acceleration [g] of the chest iPod.
  • chest_y: Y axis acceleration [g] of the chest iPod.
  • chest_z: Z axis acceleration [g] of the chest iPod.
  • waist_x: X axis acceleration [g] of the waist sensor.
  • waist_y: Y axis acceleration [g] of the waist sensor.
  • waist_z: Z axis acceleration [g] of the waist sensor.
  • right_x: X axis acceleration [g] of the wrist sensor.
  • right_y: Y axis acceleration [g] of the wrist sensor.
  • right_z: Z axis acceleration [g] of the wrist sensor.

Label data

Label data files each of which corresponds to the sensor data files.
Since the action_names were recorded in Japanese originally, they are left as they were.

Example of a CSV file:

action_id,action_name,start_t,finish_t
41,看護記録,2014-02-19 08:49:29,2014-02-19 08:49:35
,身の周りの世話,2014-02-19 08:52:07,2014-02-19 08:52:29
,身の周りの世話,2014-02-19 08:52:57,2014-02-19 08:53:31
41,看護記録,2014-02-19 08:53:44,2014-02-19 08:53:50
,身の周りの世話,2014-02-19 08:53:57,2014-02-19 08:54:28
,身の周りの世話,2014-02-19 08:54:38,2014-02-19 08:55:16
,報告連絡相談調整,2014-02-19 08:55:46,2014-02-19 08:56:49
25,心電モニター除去,2014-0,2-19 08:56:06,2014-02-19 08:56:41
,報告連絡相談調整,2014-02-19 08:57:11,2014-02-19 08:58:06
......

Each column is as follows:

  • action_id: The activity ID which corresponds to the action_id in acts.csv. We also recorded the activity classes which are predefined in acts.csv. In such cases, the action_id is empty and freely written in the next action_name.
  • action_name: The activity class name (in Japanese). If only the predefined activity classes are used, just ignore this column ( and the rows with empty action_ids.)
  • start_t: the start time of the activity.
  • end_t: the end time of the activity.

Note: If there are concurrent activities, each of them were recorded in separate lines with separate start/end times.

Duty data for unlabelled sensor data ( duties.csv )

This is the nurses' duty records extracted from the hospital information system.

Example of a CSV file:

"duty_id","date","nurse_id","start","finish" 
1,2012-02-22,3,2012-02-22 09:00:00,2012-02-22 17:00:00
2,2012-07-21,3,2012-07-21 09:00:00,2012-07-21 17:00:00
3,2011-12-09,34,2011-12-09 09:00:00,2011-12-09 17:00:00
4,2012-02-16,5,2012-02-16 17:00:00,2012-02-17 09:00:00
5,2012-07-19,5,2012-07-19 17:00:00,2012-07-20 09:00:00
6,2011-11-25,32,2011-11-25 17:00:00,2011-11-26 09:00:00
7,2011-12-09,32,2011-12-09 17:00:00,2011-12-10 09:00:00
8,2011-12-25,32,2011-12-25 17:00:00,2011-12-26 09:00:00
9,2011-12-25,32,2011-12-25 17:00:00,2011-12-26 09:00:00
......

Each column is as follows:

  • duty_id: unique id of the row.
  • date: the date (YYYY-MM-DD) the duty of a nurse.
  • nurse_id: integers to identify nurses.
  • night: TRUE if the duty is night shift, and FALSE otherwise.
  • start: start datetime (YYYY-MM-DD hh:mm:ss) of the duty. It is set to 17 pm if the night shift is TRUE, and 9 am otherwise.
  • finish: finish datetime (YYYY-MM-DD hh:mm:ss) of the duty. It is set to 9 am if the night shift is TRUE, and 17 pm otherwise.

Unlabelled sensor data

Sensor data files each of which corresponds to the duty row of the duties.csv.
The 3 types of sensor data (iPod on the chest, accelerator on the waist, and accelerator on the wrist) were merged to a one file.

Example of a CSV file:

"time","chest_x","chest_y","chest_z","waist_x","waist_y","waist_z","right_x","right_y","right_z" 
0,0.02388,-0.61212,-0.74812,-0.21875,0.90625,0.03125,-0.329411764705883,0.298039215686274,-0.831372549019608
0.05,-0.14136,-0.59595,-0.84737,-0.21875,0.953125,0.0625,-0.329411764705883,0.329411764705882,-0.831372549019608
0.1,-0.14339,-0.43019,-0.74637,-0.21875,0.921875,0.03125,-0.329411764705883,0.329411764705882,-0.831372549019608
0.15,-0.2749,-0.41902,-0.81593,-0.234375,0.9375,0.015625,-0.329411764705883,0.329411764705882,-0.831372549019608
0.2,-0.01614,-0.37914,-0.98813,-0.234375,0.9375,0.03125,-0.329411764705883,0.329411764705882,-0.831372549019608
0.25,-0.22969,-0.47845,-0.76382,-0.234375,0.921875,0.015625,-0.329411764705883,0.329411764705882,-0.831372549019608
0.3,0.12611,-0.31467,-0.82042,-0.1875,0.96875,0.03125,-0.329411764705883,0.329411764705882,-0.831372549019608
0.35,-0.04008,-0.25723,-0.96201,-0.234375,0.9375,0.03125,-0.329411764705883,0.329411764705882,-0.831372549019608
0.4,-0.07266,-0.36247,-0.85838,-0.203125,0.96875,0.046875,-0.329411764705883,0.329411764705882,-0.831372549019608
.....

Each column is as follows:

  • time: relative time [seconds] from the start time of the #{duty_id} of the duties.csv (can be identified by the file name).
  • chest_x: X axis acceleration [g] of the chest iPod.
  • chest_y: Y axis acceleration [g] of the chest iPod.
  • chest_z: Z axis acceleration [g] of the chest iPod.
  • waist_x: X axis acceleration [g] of the waist sensor.
  • waist_y: Y axis acceleration [g] of the waist sensor.
  • waist_z: Z axis acceleration [g] of the waist sensor.
  • right_x: X axis acceleration [g] of the wrist sensor.
  • right_y: Y axis acceleration [g] of the wrist sensor.
  • right_z: Z axis acceleration [g] of the wrist sensor.
Note:
  1. The time synchronization step was not applied for the unlabelled data.
  2. The sensor data has "NA" values when it was unavailable.
  3. there are several files which lacks some columns when whole values were unavailable for the duration.

Notes

The data have been collected together with the data of RFID tag data to recognize entrees of the nurses to entry the patients’ rooms, hospitalized patients’ vital data such as cardiogram, bed sensors to measure heart-rate/breath/body-movement, accelerometer, in-room sensors, and also medical information which were recorded in the electronic clinical pathways and indirectly in patients’ sensor data.
We will provide such additional data combined with this dataset as soon as getting ready.

Members

管理者: Inoue Sozo