Catalog of Regulatory Science Tools to Help Assess New Medical Devices
This regulatory science tool presents an open-access dataset tool for the assessment of wearable technology deriving gait parameters.
Technical Description
This open-access dataset tool consists of time-synchronized smartphone inertial measurement unit (IMU), reference IMU, and pressure-sensing walkway data collected from healthy participants as they walked in straight-line and curved trajectories with two different smartphones placed at varying positions and orientations on the body.
Smartphone IMU data include 3-axis acceleration and 3-axis gyroscope signals from each of two smartphones (iPhone 10 (iPhone) and Samsung Galaxy S22 (Android)); reference IMU data include 3-axis acceleration and 3-axis gyroscope signals from each of two reference IMUs (Xsens MTw Awinda); pressure-sensing walkway data include one set of binary contact data for left and right foot contact on the walkway (Protokinetics Zeno Walkway 2’ x 16’). The dataset, as collected, contained 400 CSV files representing 20 healthy participants x 2 smartphone placements x 5 smartphone orientations x 2 gait trial types. The sampling frequency of the IMU sensors and walkway were set to 100 Hz. The dataset can be accessed under Supporting Documentation, along with details on the experimental protocol, data synchronization and alignment approaches, and recorded data structure.
Intended Purpose
This dataset can be utilized to elucidate the effect of smartphone positioning on resultant smartphone IMU data relative to a reference IMU system and walkway data. In addition, the data provided can be utilized for other gait algorithm development and validation purposes.
Many wearable devices are intended to be used by the general population or patients outside a clinical environment, but the ability of the patient/user to correctly place the devices on the body has not been thoroughly evaluated and the variability introduced by incorrect placement of different types of smart devices capturing movement is not well characterized [1, 2]. An understanding of how different device (IMU) positions on the body and different types of smartphones impact the accuracy and reliability of the device output is necessary to adequately evaluate wearable technology [3, 4, 5].
Testing
Data from the walkway and IMUs were time-synchronized using a TTL-pulse such that the walkway and IMU systems started recording data at the same time (± 0.02 seconds). After the walkway and reference IMU systems were started, synchronization of these two systems with each smartphone was achieved by holding the corresponding reference IMU to the back of the smartphone and eliciting a high impulse event (shake) captured by both systems. This action was repeated for the other reference IMU + smartphone pair. Prior to alignment, the sensor coordinates systems of each device (iOS smartphone, Android smartphone, and reference IMU) were transformed such that all coordinate systems were consistent. Data alignment was achieved by time-matching the highest peak of the total acceleration signal generated from the intentional high impulse event occurring at the beginning of each trial across IMUs.
All data trial files were reviewed for completeness and quality. Due to acquisition errors, 26 trials were discarded, leaving a total of 374 trials in the final published dataset.
Limitations
Although the dataset tool provides data from two different types of smartphones (iOS and Android), specific phones using each of the listed operating systems had to be selected. To mitigate this limitation, uncalibrated data recorded directly from the inherent phone IMU sensors are reported. However, users of the tool should consider differences in the smartphone size compared to those used in the current dataset as well as operating system updates and changes to sensor technology that impact sensor recordings when determining applicability of the tool for their use.
The dataset tool also limits the number of positions on the body from which data are collected to the lower back and right thigh, thus the dataset is only applicable to technologies and algorithms using these locations.
There are currently no peer-reviewed publications describing or using the dataset, but Wiki pages have been created on the public repository to describe the data, thus enabling the immediate release of the dataset tool. Peer-reviewed publications associated with the dataset will be included in the Supporting Documentation section when available.
Supporting Documentation
The dataset can be accessed below, which includes comprehensive details on the experimental protocol, methods for data synchronization and alignment, as well as the structure of the raw data Evaluating factors that impact precision and accuracy of gait metrics derived from smartphones.
References:
Alanezi K, Mishra S (2013). "Impact of smartphone position on sensor values and context discovery". Computer Science Technical Reports. Paper 1030. http://scholar.colorado.edu/csci_techreports/1030 Mourcou Q, Fleury A, Franco C, Klopic F, Vuillerme N (2015). “Performance evaluation of smartphone inertial sensors measurement for range of motion” Sensors. 15, 23168-23187. doi:10.3390/s150923168 Coscun D, Incel OD, Ozgovde A (2015). "Phone position/placement detection using accelerometer: impact on activity recognition". IEEE Tenth International Conference on Intelligent Sensors, Sensor Networks, and Information Processing (ISSNIP). doi: 10.1109/ISSNIP.2015.7106915 Incel OD (2015). "Analysis of movement, orientation and rotation-based sensing for phone placement recognition". Sensors. 14, 25474-25506. doi:10.3390/s151025474 Kuhlmann T, Garaizar P, Reips U (2020). Smartphone sensor accuracy varies from device to device in mobile research: the case of spatial orientation. Behavior Research Methods. 53, 22-33. https://doi.org/10.3758/s13428-020-01404-5
Contact
Tool Reference
- In addition to citing relevant publications please reference the use of this tool using RST24HD01.01