Physical Therapy Sensor Data Analysis Dashboard

Published December 1, 2025

Data Analytics Software QA Janaka Alwis

Introduction

The Physical Therapy Sensor Data Analysis Dashboard was developed using the Dashtera no-code analytics environment to examine wearable motion sensor data collected during physiotherapy exercises. The dataset, originally compiled by Aras Yurtman (KU Leuven) and Billur Barshan (Bilkent University), was designed to support research on the automated evaluation of physical therapy exercises using multiple wearable motion sensors. By transforming raw sensor readings into interactive visualizations, this dashboard aims to support physiotherapists, researchers, and data scientists in understanding movement patterns, comparing exercise sessions, and examining the contribution of individual sensor units to overall motion dynamics.

Dataset

The dataset was recorded using five Xsens MTx sensor units, each equipped with tri-axial accelerometers, gyroscopes, and magnetometers. These sensors captured linear acceleration, angular velocity, and magnetic field intensity during human motion at a sampling rate of 25 Hz. Ethical approval for data collection was granted by the Bilkent University Ethics Committee.

The original research associated with this dataset can be found in:
Yurtman and Barshan (2014), Computer Methods and Programs in Biomedicine, 117(2), 189–207.

The dataset consists of:

Subjects: s1–s5
Exercises: e1–e8
Sensor Units: u1–u5 (placed on upper arm, forearm, torso, thigh, and shank)

For each subject and exercise, two session types are provided:

Template Sessions: Controlled reference recordings (3 repetitions per exercise)
Test Sessions: Realistic execution recordings (10 repetitions per exercise)

Each sensor unit includes:

Sensor Type	Axes	Measurement
Accelerometer	acc_x, acc_y, acc_z	Linear acceleration (m/s²)
Gyroscope	gyr_x, gyr_y, gyr_z	Angular velocity (°/s)
Magnetometer	mag_x, mag_y, mag_z	Magnetic field strength (µT)

A review of the record distribution indicates that template sessions are smaller and more controlled, whereas test sessions contain approximately three to four times more samples. This structure provides a suitable foundation for both baseline comparison and variability analysis across exercises and subjects.

About Dashtera

Dashtera is a cloud-based, no-code analytics platform designed to support the visual exploration and analysis of complex datasets. The platform enables users to construct interactive dashboards without programming, allowing for efficient examination of multidimensional data through line plots, bar charts, maps, regressions, and statistical summaries. Its interface allows data to be filtered, compared, and inspected from multiple perspectives, which makes it suitable for exploratory data analysis tasks.

Key Features

Integration with multiple data sources, including CSV files, APIs, and external repositories.

Support for a wide range of visualization types, such as line charts, bar charts, Pareto charts, and geographic maps.

Interactive drill-down capabilities for detailed examination of specific data segments.

Dynamic filtering that enables focused analysis based on selected criteria.

Built-in options for sharing dashboards to facilitate collaborative research and analysis.

Relevance to This Project

Dashtera was selected for this study due to its ability to rapidly generate interactive, sensor-focused dashboards without additional software development overhead. Its lightweight configuration and visual workflow make it well suited for time-series comparison, motion pattern inspection, and exploratory analysis of multichannel sensor data.

Dashboards

Dataset Overview

The first dashboard provides a structural summary of the dataset across subjects, exercises, and session types.

Table 1 presents the distribution of sensor records for template and test sessions for each subject, while Table 2 shows the corresponding distribution across exercises. Table 3 summarizes total record counts.

Table 1. Record distribution by subject

Subject	Template Records	Test Records
s1	75,890	236,975
s2	52,790	212,640
s3	54,815	224,955
s4	52,305	221,005
s5	40,825	205,815

Table 2. Record distribution by exercise

Exercise	Template Records	Test Records
e1	39,845	139,615
e2	37,345	139,270
e3	36,380	139,320
e4	32,660	137,115
e5	31,795	136,190
e6	33,285	136,970
e7	33,505	136,000
e8	31,810	136,910

Table 3. Total record counts

Exercise	Template Records	Test Records
All Data Combined	276,625	1,101,390

The record distributions demonstrate that template sessions are smaller and more controlled, whereas test sessions contain approximately three to four times more samples. Sensor units have identical record counts across subjects, confirming consistent device recording. This balance ensures that subsequent analyses reflect true motion variations rather than sampling inconsistencies.

Key Insight: Template data provide a controlled reference baseline, while test sessions capture variability across repeated executions, supporting comprehensive comparative and variability analyses.

Template Session Analysis (Subject 1 · Exercise 1)

Template session dashboards for subject 1 for exercise 1. display synchronized accelerometer, gyroscope, and magnetometer time-series signals for each sensor unit.

For each subject and exercise, these plots exhibit smooth periodic patterns with minimal noise, indicating consistent execution. Peaks in accelerometer readings align closely with gyroscope rotations, reflecting coordinated limb motion. Magnetometer trajectories further demonstrate stable orientation, corroborating the controlled nature of template sessions.

Key Observations:

Signals are smooth, periodic, and low-noise.
Gyroscope peaks are aligned with accelerometer maxima, confirming temporal synchronization of rotational and linear motion.
Magnetometer readings remain consistent, indicating stable orientation.

These template signals serve as a baseline for subsequent comparisons with test sessions, enabling the identification of deviations from controlled reference movement.

Test Session Analysis (Subject 1 · Exercise 1)

Test session dashboards for subject 1 and exercise 1 exhibit increased variability in amplitude, frequency, and noise compared to template sessions.

Each subject performed 10 repetitions per exercise, introducing natural fluctuations in movement speed, strength, and rhythm. Visual inspection reveals both inter-subject and intra-subject variability, particularly in distal segments such as the forearm and wrist. Compared with template sessions, higher amplitude and irregular patterns in accelerometer and gyroscope signals reflect realistic exercise conditions.

Key Observations:

Amplitude variation highlights differences in execution speed and strength.
Increased signal noise corresponds to natural variability in human performance.
Useful for evaluating execution consistency and identifying outlier repetitions.

3D Motion Visualization (Subject 1)

For subject 1, 3D trajectory dashboards plot sensor data across x, y, and z axes for each sensor unit, providing a spatial representation of limb and body movement.

Accelerometer trajectories illustrate cyclic motion patterns, gyroscope plots highlight angular rotation and spread, and magnetometer plots reveal orientation dynamics and environmental field response. These 3D visualizations complement time-series analyses by enabling geometric interpretation of kinematic behavior.

Key Observations:

Accelerometer trajectories form rhythmic, cyclic paths corresponding to repetitive limb motion.
Gyroscope plots reveal directional rotation and angular spread.
Magnetometer plots capture sensor orientation relative to the environment.
Combined 3D visualizations provide a comprehensive understanding of spatial motion dynamics.

Sensor Unit Analysis (Subject 1 · Exercise 1)

This dashboard examines Subject 1 performing Exercise 1 at the sensor-unit level.

The bar charts summarize record counts per exercise and per sensor, confirming uniform data collection. Time-series plots for each unit display accelerometer, gyroscope, and magnetometer signals, illustrating linear, rotational, and orientation dynamics.

Key Observations:

Accelerometer signals reflect linear displacement, gyroscope signals capture rotational activity, and magnetometer signals indicate orientation stability.
Distal units, such as the forearm, exhibit more pronounced and rapid motion, whereas torso units show minimal displacement.
The combined modalities provide a comprehensive view of segment-specific kinematics, supporting subsequent inter-sensor comparisons.

Sensor Unit Analysis (Subject 1 · Exercise 1 · Unit 2)

Unit-level dashboards allow examination of motion contributions from individual body segments. Table 4 summarizes sensor placement and motion characteristics.

Unit	Placement	Motion Characteristics
u1	Upper Arm	Large amplitude, slower motion
u2	Forearm/Wrist	High rotational velocity, rapid transitions
u3	Chest/Torso	Minimal motion, orientation stability
u4	Thigh	Moderate leg motion
u5	Shank/Ankle	Strong acceleration, high frequency

Key Observations:

u2 (Forearm/Wrist) provides the clearest and most informative signals due to rapid motion and high rotational velocity.
u3 (Torso) shows minimal displacement, capturing primarily postural stability.
Signals from u1, u4, and u5 correspond to expected segment-specific biomechanical behavior.
Combined, these data allow segment-level analysis of limb dynamics and stability.

Inter-Sensor Relationship Analysis (s1 · e1 · u2)

Inter-sensor scatterplots illustrate relationships between accelerometer, gyroscope, and magnetometer measurements along each axis. Table 5 summarizes key relationships.

Sensor Pair	Relationship	Description
acc_x ↔ gyr_x	No linear correlation	Linear vs. angular measurement domains are independent
acc_x ↔ mag_x	Negative linear	Increased linear acceleration slightly affects orientation
gyr_x ↔ mag_x	Circular scatter	Rotational and magnetic signals largely independent
acc_y ↔ gyr_y	No clear relation	Orthogonal measurement domains
acc_y ↔ mag_y	Negative trend	Minor acceleration–orientation interaction
acc_z ↔ gyr_z	Slight curve	Vertical rotation coupling evident
acc_z ↔ mag_z	Negative slope	Opposite field direction with acceleration
gyr_z ↔ mag_z	Circular	Angular rotation and magnetic field axes largely independent

Interpretation:

Accelerometer–Magnetometer pairs exhibit directional correlation, suggesting that linear motion influences magnetic orientation.
Other sensor pairs are largely uncorrelated, indicating that each modality captures distinct physical aspects: linear, rotational, and orientation dynamics.
These observations confirm proper sensor calibration and support multimodal data fusion in further analysis.

Discussion

The dashboards collectively demonstrate that the dataset is well structured for analyzing physiotherapy exercise quality. Template sessions provide clean kinematic baselines, while test sessions offer realistic variability useful for performance evaluation or machine learning training. Sensor relationships and unit-wise differences reveal expected biomechanical patterns and confirm proper device calibration.

Conclusion

The Physical Therapy Sensor Data Analysis Dashboard demonstrates the effectiveness of Dashtera as an interactive, no-code platform for analyzing complex motion sensor data. By leveraging Dashtera’s visualization capabilities, users can:

Examine kinematic characteristics at multiple levels, including session, sensor unit, and individual axes.
Compare controlled template exercises with test executions to assess variability.
Identify meaningful inter-sensor relationships across linear, rotational, and orientation measurements.
Explore and interpret movement patterns without requiring programming expertise.

Implications:

Dashtera enables physiotherapists, biomechanics researchers, and data scientists to evaluate motion quality, analyze segment-specific dynamics, and detect deviations from reference exercise patterns efficiently. Its interactive dashboards, combined with a structured, high-resolution dataset, provide a robust foundation for computational modeling, feature extraction, and the development of AI-assisted motion classification or rehabilitation assessment tools.

Summary:

The integration of high-resolution motion sensor data with Dashtera’s interactive, no-code environment transforms complex kinematic signals into actionable insights. By facilitating rapid exploration, comparison, and visualization of human movement, Dashtera significantly enhances both research and applied assessment in physical therapy, supporting data-driven rehabilitation strategies and biomechanical analysis.

Category: