PerHeart Pilot Dataset: Wrist-worn Inertial Sensor and Physiological Data from Older Adults with Heart Failure

1 PerHeart Pilot Dataset: Wrist-worn Inertial Sensor and Physiological Data from Older Adults with Heart Failure Authors: Marcin Kolakowski1, Vitomir Djaja-Josko1, Jerzy Kolakowski1, Irina Mocanu2,3, Oana Cramariuc3, Jerzy Gasowski4, Karolina Piotrowicz4, Affiliations: 1 Institute of Radioelectronics and Multimedia Technology, Warsaw University of Technology, Warsaw, Poland 2 Computer Science Department, Faculty of Automatic Control and Computers, National University of Science and Technology Politehnica Bucharest, Bucharest, Romania 3 Centrul IT Pentru Stiinta si Tehnologie (CITST), Bucharest, Romania 4 Department of Internal Medicine and Gerontology, Jagiellonian University Medical College, Cracow, Poland Abstract: This dataset contains acceleration and angular velocity measurement results from wrist-worn sensors and physiological data collected using medical devices (blood pressure meter, pulse oximeter, thermometer, bathroom scale, and glucometer) during a pilot study of the PerHeart platform in Poland. It includes data from 27 older adults with heart failure history who participated in one-month long trials. Eight adults’ activities were measured using inertial sensors resulting in 2,536 hours of acceleration and angular velocity. The dataset also provides step count data (over 687,000 steps detected) and barometric pressure readings. These data can support research on daily activity patterns and gait analysis in older individuals with heart failure, and are well-suited for machine learning applications, including semi-supervised learning scenarios using unlabeled time-series data. DOI: https://www.doi.org/10.5281/zenodo.17143200 Keywords: wearable devices, heart failure, older adults, gait analysis, physical activity, time-series data, semisupervised learning, medical monitoring 2 1. Background and summary The main aim of the PerHeart (Personalized ICT solution to reduce re-hospitalization rates in heart failure elderly patients suffering from comorbidities) project was to develop a healthmonitoring platform, which would help to reduce hospitalization rates of the people suffering from heart-related problems. The platform integrated commonly used medical devices (incl. blood pressure meters, bathroom scales) as well as custom made sensors – intelligent pillbox and a wrist worn activity/gait sensor. At the start of the pilots, the participants used Mi Band 3 smart bands. However, due to difficulties with obtaining raw measurement data, they were replaced with custom made wrist sensors at Warsaw University of Technology. The new sensors included an inertial measurement unit and a barometer constantly recording data during the user’s daily routines. The dataset includes data collected over the pilot study from 27 (8 of which have worn the WUT wrist sensor) older adults suffering from heart failure. All the pilot participants were patients of University Hospital in Cracow, Poland. All participants signed written consent forms. The study was accepted by the Jagiellonian University Ethics Committee with reference 1072.6120.17.2023 (date: 15/02/2023). 2. Collection methods The dataset contains the data collected during the pilots performed in Poland. In the pilots eight older adults suffering from heart failure were recruited and used a system for a minimum of one month. Each of the participants was supplied with the following set of devices: • a tablet, • a digital bathroom scale (UC-352BLE), • a digital blood pressure meter (UA-651BLE or BM 95), • a digital thermometer (UT-201BLE-A), • a digital pulse oximeter (Jumper JPD-500F BLE), • a digital glucometer (OneTouch Select Plus Flex) used by selected users, • an intelligent pillbox developed by CITST, • a wrist-worn sensor developed by WUT. The tablets were preinstalled with custom PerHeart applications which allowed for collecting data from the rest of the sensors and visualizing them in a processed form so the users could monitor their status. The medical devices communicated with the tablet using Bluetooth, whereas the wrist-worn sensor relied on an USB connection through a specially prepared charging interface. The tablet application relayed the results to the PerHeart database without any processing, so the dataset contains raw measurement results collected with the devices. To obtain values in the standard units (e.g. g for acceleration), the values should be converted based on the measurement ranges listed in Table 1. 3 The wrist-worn sensors include a BLE-enabled nRF52833 Nordic Semiconductor microcontroller, one Bosch Sensortec BMI270 Inertial Measurement Unit [1] comprising a 16bit tri-axial gyroscope and a 16-bit tri-axial accelerometer and one BMP390 barometer [2]. They are encased in Mi Band strap compatible 3D printed cases. Each device is equipped with a small 220 mAh battery which allows for over two days of constant operation. The direction of IMU axes in respect to the sensor case is presented in Figure 1. Figure 1. Orientation of sensor axes. The hand on which the sensor is worn was not enforced The device's sensors operate with predefined sampling rates and resolutions which were constant throughout the whole pilot study and are listed in Table 1. Table 1 Sensor parameters Sensor Sampling rate Range Sensitivity IMU (acceleration) 50 Hz ±4g LSB/g IMU (angular velocity) 50 Hz ± 2000 dps 16.384 LSB/dps IMU (internal step counter) 1.25 Hz - 1 step Atmospheric pressure meter 2.5 Hz - 0.17 Pa1 1 Averaged based on 16 last measurements taken with 200 Hz rate 4 3. Data records The dataset is available in the Zenodo repository at https://www.doi.org/10.5281/zenodo.17143200. The directory tree of the dataset is presented below: └── Dataset/ ├── wrist/ │ ├── user_1/ │ │ ├── user_1_file_01_inertial_1700980027.parquet │ │ ├── user_1_file_01_pressure_1700980027.parquet │ │ ├── user_1_file_01_steps_1700980029.parquet │ │ └── ... │ ├── user_2/ │ │ ├── user_2_file_01_inertial_1698074792.parquet │ │ ├── user_2_file_01_pressure_1698074792.parquet │ │ ├── user_2_file_01_steps_1698074794.parquet │ │ └── ... │ └── ... ├── medical/ │ ├── blood_pressure.csv │ ├── body_mass.csv │ ├── glucose.csv │ ├── oxidation.csv │ └── temperature.csv ├── personal_questionnaires.csv ├── load_dataset.ipynb ├── load_dataset.py └── perheart_dataset_descriptor.pdf The measurement results and user demographics information are stored in common csv and parquet formats. The basic information on the users is stored in the personal_questionnaires.csv file. The accompanying results from the medical devices are grouped by measurement type in files located in medical directory. The inertial measurements are stored in the /wrist directory. The inertial measurement files are named using the convention: user_{user_id}_file_{file_no}_{type}_{ts}.parquet, where: • user_id – user identifier as in Table 2 (from 1 to 8), • file_no – file number (does not correspond to day of the pilot), • type – file type, either ‘inertial’, ‘pressure’ or ‘steps’, • ts – timestamp in seconds since epochs (UTC, measurements were conducted in Poland which is UTC+1 or UTC+2 for the measurements taken before Oct 29th 2023 so it should be taken into consideration when analyzing daily routines). Both medical and wrist directories were compressed to save space and make the repository satisfy the Zenodo file number limit 5 3.1 Personal questionnaires The demographics files include the answers to the personal questionnaire filled in before and after the pilot study, the MMSE score and the Barthel ADL Index. The questions and answers are stored in Table 2. Table 2. Personal questionnaire variables Question Parameter name values user identifier user_id a unique user identifier (integer from 1 to 27) Date of start start_date Date of start MMSE mmse_total Total MMSE score 1. age age age in years 2. sex sex 1 - male, 2female 3. What are your current living arrangements? living 1 - alone 2 - with a partner 3 – recently alone 4 - with professional support 4. What is your current job situation? job 1 – retired 2 – long term leave due to illness or disability 3 – employed 4 – unemployed 5. How often do you leave your home in a typical week? outdoor_freq 1 – (almost) every day 2 – two to four days a week 3 – once a week or less 6. How satisfied are you when it comes to your social situation? social_satisfaction 1 – totally dissatisfied 2 – quite satisfied 3 – satisfied 4 – very satisfied 5 – exceptionally satisfied 7. How satisfied are you when it comes to your financial situation? financial_satisfaction 1 – totally dissatisfied 2 – quite satisfied 3 – satisfied 4 – very satisfied 5 – exceptionally satisfied 8. Do you own a smartphone? smartphone 0 – no, 1 – yes 9. Do you own any other devices other_devices 0 – no 1 – PC 2 – laptop 3 – tablet 10. How knowledgeable are you when it comes to using technical devices? digital_proficiency 1 – not at all 2 – quite knowledgeable 3 – knowledgeable 4 - very knowledgeable 5 – exceptionally knowledgeable 6 11. How easy is it for you to start using new devices? new_devices 1 – not easy 2 – quite easy 3 – easy 4 – very easy 5 – exceptionally easy 12a. I think that it is nice to use new technological gadgets. tech_1 1 – strongly disagree 2 – disagree 3 – undecided 4 – agree 5 – strongly agree 12b. Using technology makes my life easier tech_2 1 – strongly disagree 2 – disagree 3 – undecided 4 – agree 5 – strongly agree 12c. I like to buy the newest models and updates. tech_3 1 – strongly disagree 2 – disagree 3 – undecided 4 – agree 5 – strongly agree 12d. I sometimes worry that I will not be able to use new technical solutions. tech_4 1 – strongly disagree 2 – disagree 3 – undecided 4 – agree 5 – strongly agree 12e. The current technological development is so fast that it is hard to follow. tech_5 1 – strongly disagree 2 – disagree 3 – undecided 4 – agree 5 – strongly agree 12f. I would like to try to use new gadgets to a higher degree if I had more support and help than now. tech_6 1 – strongly disagree 2 – disagree 3 – undecided 4 – agree 5 – strongly agree 13a. How serious were the heart failure consequences? hf_1 1 – I was able to cope on my own 2 – I needed professional help 3 – I was hospitalized 13b. When did you have the last heart failure episode? hf_2 Date in string – month/year or approximate e.g. “Summer 2023” if the patient could give only a rough estimate. 7 14. Please rate your health state. hs_1 1 – very bad 2 – bad 3 – neutral 4 – good 5 – very good 15. How satisfied are you with your health state? hs_2 1 – totally dissatisfied 2 – quite satisfied 3 – satisfied 4 – very satisfied 5 – exceptionally satisfied 16. Who helps you in your everyday life? help_1 1 – family members 2 – professional caregivers 3 – friends / neighbors 4 – I do not need help 17. Do you need help while walking? help_2 1 – Yes, walker 2 – Yes, wheelchair 3 – Yes, crutches 4 – I do not need any devices 18. How satisfied are you with the project? project_1 1 – totally dissatisfied 2 – quite satisfied 3 – satisfied 4 – very satisfied 5 – exceptionally satisfied 19a. How well do you know the PerHeart system components? [Tablet] project_2a 1 – not at all 2 – quite knowledgeable 3 – knowledgeable 4 - very knowledgeable 5 – exceptionally knowledgeable 19b. How well do you know the PerHeart system components? [Application]. project_2b 1 – not at all 2 – quite knowledgeable 3 – knowledgeable 4 - very knowledgeable 5 – exceptionally knowledgeable 19c. How well do you know the PerHeart system components? [Measurement devices]. project_2c 1 – not at all 2 – quite knowledgeable 3 – knowledgeable 4 - very knowledgeable 5 – exceptionally knowledgeable 20. Please rate the instructions and manuals received before the pilot’s start. project_3 1 – totally insufficient 2 – quite sufficient 3 – sufficient 4 – exhaustive 5 – very exhaustive 8 21. Which devices and applications did you use during the pilot? 21a. calendar project_4a 0 – no, 1yes 21b. pillbox project_4b 0 – no, 1yes 21c. thermometer project_4c 0 – no, 1yes 21d. pulse oximeter project_4d 0 – no, 1yes 21e. glucometer project_4e 0 – no, 1yes 21f. blood pressure meter project_4f 0 – no, 1yes 22. Who helped you with using the PerHeart set? project_5 1 – family members 2 – friends 3 – caregivers 4 – nurses 5 – I did not need help 23. How much did participation in the project increase your digital literacy? project_6 1 – not at all 2 – a little bit 3 – did not change anything 4 – improved 5 – significantly improved 24. How did your feeling of autonomy change during the project? project_7 1 – it’s worse 2 – it’s a bit worse 3 – no change 4 – a bit better 5 – better 25. How much did your safety perception change during the project? project_8 1 – it’s worse 2 – it’s a bit worse 3 – no change 4 – a bit better 5 – better 26a. If you experienced a recent HF, when did it happen? recent_hf_1 Date in string day/month/year 26b. If you experienced a recent HF, what did happen? recent_hf_2 Additional comment in string (may also present in case of some users who did not experience HF during the pilots). 26c. If you experienced a recent HF, were you hospitalized? recent_hf_3 0 – no, 1 – yes Barthel scale [3] barthel The Barthel ADL Index (0-100) 9 3.2 Medical measurements The measurements from the medical devices are stored in medical_{meas_type}.csv files where meas_type is one of the measurement types: blood_pressure, oxidation, body_mass, glucose, , temperature. 3.2.1 Blood pressure measurements The body temperature measurements were performed using the A&D UA-651BLE or BM 95 digital blood pressure meter. The description of the file’s structure is presented in Table 3. Table 3. Blood pressure measurements file structure variable parameter name values user identifier user_id a unique user identifier (string format) timestamp ts timestamp in seconds (Unix epoch). device model device 0 - A&D UA-651BLE 1 - Beurer BM95 systolic blood pressure bp_sys systolic blood pressure [mmHg] diastolic blood pressure bp_dia diastolic blood pressure [mmHg] heart rate hr heart rate [bpm] comment comm Additional comment: For A&D UA-651BLE - provides information on the detection of the irregular pulse - 1 if detected, 0 otherwise For Beurer BM95 - evaluation according to the WHO classification (for details see Beurer BM95 user manual [2]) 3.2.2 Blood oxidation measurements The blood oxidation measurements were performed using the Jumper JPD-500F digital pulse oximeter. The description of the file’s structure is presented in Table 4. Table 4. Blood oxidation measurements file structure variable parameter name values user identifier user_id a unique user identifier (string format) timestamp ts timestamp in seconds (Unix epoch). heart rate hr heart rate [bpm] perfusion index perf perfusion index [-] saturation sat oxygen saturation [%]