Smart Compression Sleeve for Tracking and Recovery Monitoring in Knee Rehabilitation
ABSTRACT
Rehabilitation after knee injuries, such as Anterior Cruciate Ligament (ACL) tears, is important for recovery. However, progress is often tracked using subjective visual observation, which can lead to high rates of re-injury. Motion capture systems are expensive and inaccessible for clinical use. This study aims to validate a low-cost, wearable “Smart Compression Sleeve” to provide objective, quantitative data for rehabilitation. A prototype was built using an ESP32, two BNO055 IMUs, and a flex sensor. Nine participants (five healthy, four injured) performed six standardised movement tasks. Data was processed using a Python pipeline with Madgwick fusion and Butterworth filters. The system demonstrated good reliability (e.g., 8.47% Coefficient of Variation for Angular Velocity RMS). It successfully quantified significant performance deficits in injured participants. Compared to the healthy baseline, the injured leg showed a 27.3% reduction in Knee Flexion Range and a 38.9% increase in instability metrics (Sway Path and 95% Ellipse Area). This proves the sleeve is a reliable, low-cost tool that can bridge the gap between subjective assessment and expensive labs, enabling data-driven physiotherapy.
INTRODUCTION.
The human knee is a biomechanical marvel, but its complexity makes it highly susceptible to injury. Knee injuries are serious health issues encompassing 41% of all sports injuries [1]. Among these, injuries to the Anterior Cruciate Ligament (ACL) are very common, accounting for roughly 50% of all knee ligament injuries [2]. The median return-to-sport (RTS) time following these injuries can exceed one year [3], carrying high personal and economic costs [4].
Following an ACL reconstruction, the quality of post-operative physiotherapy is important. The primary goals are to restore the knee’s range of motion (ROM), rebuild muscle strength, regain proprioception, and understand the body’s position in space[5]. Rehabilitation failures are very common, with improper recovery protocols contributing to an alarming secondary ACL injury rate, which can be as high as 23% for young athletes [2]. Most of this problem lies in how recovery is monitored.
In many clinics, a therapist assesses progress with an “eye-test” or a manual goniometer. This method is subjective and can miss subtle, high-speed movement errors, such as a slight valgus collapse (the knee turning inward). This subjectivity creates ambiguity, and research has shown that failing to meet objective discharge criteria is associated with a four-times higher risk of graft rupture [6].
At the other extreme, “gold standard” optical motion capture systems offer objective, 3D data [7]. However, these systems are extremely expensive and require a dedicated lab and trained technicians. This creates a vast measurement gap. This accessibility gap is particularly pronounced in India, where clinics often rely entirely on observation-based methods, increasing the risk of premature RTS and re-injury [8].
This research aims to bridge this gap using modern wearable technology. A “smart” wearable device, built with low-cost sensors, can provide objective, quantitative data at a fraction of the cost. This study focuses on Inertial Measurement Units (IMUs) and flex sensors. This project aims to build and test a prototype to fill this gap. We propose a “Smart Compression Sleeve” integrating two BNO055 IMUs and one flex sensor with an ESP32 microcontroller. This system samples data at 50 Hz and streams it via Bluetooth Low Energy (BLE) to a smartphone for analysis.
This study is guided by two primary research questions (RQs). The first one (RQ1) is – “Can this wearable system detect clear, quantitative differences in balance and knee movement between participants with knee injuries and healthy, non-injured participants?” The second one (RQ2) is – “How consistent and reliable are the data from the wearable system when a non-injured participant repeats the same movement tasks multiple times?”
To answer these, the study has two objectives. The first objective is to implement and validate a dual-IMU and flex sensor-based BLE system capable of transmitting stable sensor data for analysis. The second objective is to extract and compare key biomechanical metrics (e.g., Sway Path, Ellipse Area) from movement trials to quantify the differences between injured and non-injured groups.
MATERIALS AND METHODS.
This section describes the design of the “Smart Compression Sleeve” prototype, the experimental procedure for data collection, and the computational pipeline used for data analysis.
System Architecture.
The project’s objective was to create a low-cost, wearable system to capture, transmit, and record key biomechanical data from the knee joint. The system architecture was designed as a simple, wireless data-logging pipeline, as shown in Figure 1.
For sensing, two Inertial Measurement Units (IMUs) and one flex sensor, integrated into a compression sleeve, were used to continuously measure the 3D orientation of the thigh and shank, as well as the knee flexion angle.
For processing, a central ESP32 microcontroller reading data from all three sensors was used. The ESP32 samples the data at a target frequency of 50 Hz and formats the readings into a single comma-separated value (CSV) data packet. Subsequently, the ESP32’s built-in Bluetooth Low Energy (BLE) module wirelessly transmits the data packet to a nearby smartphone. A standard smartphone application (nRF Connect for Mobile) was used to receive and log these data packets, saving the complete movement trial as a CSV file. Then, the exported CSV files are processed in an offline Python analysis pipeline, where key biomechanical metrics are calculated.
Hardware Components.
The hardware was selected to balance low cost, high accuracy, and portability. An ESP32 (ESP32 Dev Kit) microcontroller was used as the central processor. This component was chosen for its sufficient processing power and, most importantly, its integrated Bluetooth Low Energy (BLE) radio, which simplifies the design. To measure 3D motion, two Adafruit BNO055 IMUs were used. The BNO055 is a 9-axis sensor that contains an accelerometer, gyroscope, and magnetometer. Its key advantage is a built-in fusion algorithm that runs on the sensor itself, correcting for errors like gyroscope “drift” and providing a stable orientation output. Two IMUs were used: one for the thigh and one for the shank. To provide a direct and robust measurement of the knee’s bending angle, a simple flex sensor (Spectra) was added. This sensor functions as a variable resistor; as it is bent, its electrical resistance changes. This change is read by an analog-to-digital (ADC) pin on the ESP32. The wearable system was powered by a standard 3.7 Volt Lithium-Ion (Li-ion) battery.
System Fabrication and Assembly.
The prototype was constructed by integrating the electronic components onto a standard nylon compression sleeve. The two BNO055 IMUs were connected to the ESP32’s I2C bus, and the flex sensor was connected to an ADC pin. Sensor placement is critical for accurate data; therefore, both IMUs were placed at predetermined locations on the leg to yield the output. IMU1 (thigh) was secured 10 cm above the patella (kneecap). IMU2 (shank) was secured 10 cm below the patella. The flex sensor was placed directly across the side of the knee joint.
All components were affixed to the sleeve using a combination of hot glue and strong adhesive to ensure they would not move during the movement trials. Figure 2 is an image of the sleeve for easy reference.
Experimental Protocol.
A study was conducted to answer the research questions.
Participant Recruitment and Ethics.
A total of nine participants were recruited using convenience sampling. The cohort consisted of a Baseline Group (n=5), comprising five healthy participants with no history of knee injury, and an Injured Group (n=4), comprising four participants who are currently undergoing rehabilitation for knee injuries.
This study was conducted with strict ethical considerations. Informed verbal consent was obtained from all participants (and their guardians, if a minor). All data was anonymised.
Study Procedure.
Each data collection session lasted approximately 20 minutes. After the sleeve was worn, the researcher connected to the device using the nRF Connect app. The participant was asked to perform six specific movement tests such as 10m Jog, Double Step Stair Climbing (walking upstairs, skipping every second step), Triple Step Stair Climbing (walking upstairs, skipping every third step), Up Down Stair Climbing (climbing up and down stairs normally), Leg Sway (standing on one leg, swaying the other like a pendulum), Knee Up (standing on one leg, folding the other knee to the chest). Figure 3 shows the participants performing the tests in real-time. A 3-second stationary calibration period was recorded before and after each task.
Data Analysis Pipeline.
All data analysis was performed post-processing using a custom Python pipeline. First, raw log files were cleaned using a pre-processing script to extract the valid 20-column sensor data.
The main analysis pipeline then executed a multi-stage process for each trial. Firstly, a Madgwick fusion algorithm was implemented to fuse the 9-axis data (accelerometer, gyroscope, magnetometer) from each IMU. This converted the raw sensor readings into stable 3D orientation data (Euler angles). Secondly, a 4th-order, 6Hz Butterworth low-pass filter was applied to all sensor signals (Euler angles and flex sensor data). This removed noise and isolated the true physiological movement.
From the clean, filtered data, four key biomechanical metrics were calculated. Knee Flexion Range (deg) represented the total range of motion derived from the flex sensor. Sway Path (deg), which quantified postural stability, was calculated from the 2D roll-pitch trajectory of the thigh IMU (a higher value means more sway). 95% Ellipse Area (deg²) was a second measure of balance, representing the area of a 95% confidence ellipse enclosing the roll-pitch data (a larger area mean less control). Finally, Angular Velocity RMS (deg/s) was used. It represents the Root Mean Square of the gyroscope data, quantifying movement “shakiness”. A higher value means less smooth movement.
RESULTS AND DISCUSSION.
This section presents and interprets the quantitative findings from the experimental study. The results are organised to first establish the system’s reliability (addressing RQ2) and then to present the primary findings on patient performance (addressing RQ1).
The data was processed using the Python pipeline (as described in Materials and Methods). This analysis compares three distinct participant groups, namely Baseline (n=5), a control group of five healthy participants; Injured LL (n=2), the unaffected, “good” leg of two injured participants; and Injured RL (n=2), the recovering, “bad” leg of two injured participants.
Four key biomechanical metrics were calculated for each trial: Knee Flexion Range (deg), Sway Path (deg), 95% Ellipse Area (deg²), and Angular Velocity RMS (deg/s).
Result 1: System Reliability (RQ2).
Before assessing patient data, it was necessary to validate the reliability of the prototype. Research Question 2 asked if the system could produce consistent data across repeated trials. To answer this, data from the verification subject (a non-injured participant who performed the protocol on three separate days) were analysed. We also assessed the intra-group variability using the Coefficient of Variation (CV%).
The system demonstrated good reliability. For example, during the “Double Step Stair Climbing” task, the Coefficient of Variation for Angular Velocity RMS within the baseline group was only 8.47%. This value is well below the 15% threshold that is typically considered “good” reliability in biomechanical studies. This low variability confirms that the smart sleeve is a consistent and trustworthy measurement tool. This result is crucial. It gives us confidence that the differences reported in the next section are a true reflection of physiological differences between the participants, not random sensor error.
Result 2: Principal Findings (RQ1).
The primary goal of this study (RQ1) was to determine if the smart sleeve could detect clear, quantitative differences between injured and non-injured individuals. The results provided a clear and affirmative outcome. Across all six movement tasks, from a simple “10m Jog” to a challenging “Triple Step Stair Climbing” test, a consistent, hierarchical pattern of performance emerged. As expected, the baseline group exhibited superior performance to the injured leg group. The healthy leg of the injured participants (the left leg) performed at an intermediate level. This deficit was not only consistent but also quantifiable. The analysis revealed two primary findings: a 27.3% reduction in range of motion and a 38.9% increase in markers of instability.
Deficit 1: Knee Flexion Range (ROM).
A primary goal of rehabilitation is to restore the knee’s full range of motion. The data showed a significant deficit in the injured group. Compared to the Baseline (198.00 deg), the injured leg (RL) group achieved an average of only 144.00 deg of flexion. This is a 27.3% reduction in ROM.
This finding was consistent across all six activities. Figure 4 illustrates this deficit during the “Triple Step Stair Climbing” task. The chart clearly shows the step-down in performance from the Baseline group to the injured leg.
This 27.3% deficit is a quantitative measure of clinical symptoms. A reduced ROM is a hallmark of post-operative recovery. It is caused by factors such as joint swelling (edema), post-surgical stiffness (arthrofibrosis), and “muscle guarding”, an involuntary tensing of the muscles to prevent pain. The smart sleeve successfully translates these complex symptoms into a single, objective number.
Deficit 2: Postural Instability and Movement Control.
The second major finding was a severe loss of balance and control in the injured group. This was quantified by three separate metrics: Sway Path, 95% Ellipse Area, and Angular Velocity RMS.
In every task, the injured leg (RL) group showed a 38.9% increase in these metrics compared to the healthy Baseline. This finding is perhaps more significant than the flexion deficit. An ACL tear is not just a mechanical injury; it is a neurological one. The ACL is rich with mechanoreceptors, nerves that send proprioceptive signals (the sense of joint position) to the brain. When the ACL is torn, this signal is lost. The brain must then rely more on vision and muscular control to stay balanced. The +38.9% increase in Sway Path and Ellipse Area is the quantitative evidence of this proprioceptive loss. The participant’s body is “hunting” or “swaying” more to find its centre of balance. Figure 5 illustrates this clearly. During the “Leg Sway” balance test, the area of sway for the injured leg is dramatically larger and less controlled than the baseline.
This instability was further confirmed by the +38.9% increase in Angular Velocity RMS. This metric quantifies the “shakiness” of a movement. The injured group’s movements were jerkier and less smooth, reflecting a lack of fine neuromuscular control. Finally, an interesting finding was the “bilateral deficit.” The non-injured leg (LL) of the injured participants consistently performed worse than the healthy baseline. This is a known phenomenon where individuals adopt a compensatory movement strategy to protect their injured side, which affects both legs. The smart sleeve was sensitive enough to detect even this subtle, intermediate level of impairment.
Discussion.
The findings of this study are significant, as we have successfully designed, built, and tested a low-cost prototype that is both reliable (addressing RQ2) and valid (addressing RQ1). This work builds directly on the foundational literature that shows IMUs can measure knee kinematics. Our study takes the next logical step: we applied this technology to a clinical problem and successfully used it to distinguish between healthy and injured populations. This project directly addresses the “measurement gap” identified in the Introduction. Rehabilitation monitoring is currently caught between the subjective, unreliable “eye-test” of a therapist and the prohibitively expensive “gold standard” Vicon lab. Our “Smart Compression Sleeve” is a practical solution that bridges this gap.
The significance of this, particularly in the context of economic disparities in India, is high. This is a tool that could be used by local physiotherapy clinics that cannot afford expensive lab equipment. It empowers a physiotherapist to make data-driven decisions. Instead of saying, “Your balance looks better,” they can say, “Your 95% Ellipse Area has decreased by 30%.” This quantitative approach is essential for preventing re-injury.
CONCLUSION.
This research successfully demonstrated that a low-cost, sensor-based “Smart Compression Sleeve” can provide reliable and quantitative data for knee rehabilitation. We addressed our primary objectives by proving the system’s reliability (RQ2), with low Coefficients of Variation (e.g., 8.47% CV). We also confirmed our main hypothesis (RQ1) by detecting clear, multi-dimensional performance deficits in injured participants. Our findings showed the injured leg had a 27.3% reduction in flexion range and a 38.9% increase in instability metrics (Sway Path and Ellipse Area). This successfully quantifies the loss of both motion and postural control. This study’s key contribution is a validated, accessible tool that bridges the gap between subjective clinical guesswork and expensive, lab-based systems. Limitations include a small sample size (n=9), the lack of gold standard validation, and the prototype’s non-robust design. Future work will focus on hardware miniaturisation onto a PCB, validation against a Vicon system, and development of a user-friendly app. This study is a promising proof-of-concept for a tool that can make objective, data-driven knee physiotherapy accessible to all.
ACKNOWLEDGEMENTS.
The author thanks Abirami Rajasekaran, Sankar Balasubramanian, and Aashna Saraf for their guidance and mentorship throughout the development of the project.
REFERENCES.
- Golberg et al., Wearable sensors and smart devices to monitor rehabilitation in athletes and post-operative patients. Sensors 23, 1234–1256 (2023).
- Tan et al., A scoping review of portable sensing for out-of-lab anterior cruciate ligament injury prevention and rehabilitation assessment. NPJ Digit. Med. 6, 89–104 (2023).
- Wu et al., Monitoring external workload with wearable technology after ACL reconstruction: A systematic review. Am. J. Sports Med. 51, 1123–1135 (2023).
- Guan et al., Validity of wearable sensors for total knee arthroplasty rehabilitation: A systematic review. Knee 45, 78–90 (2024).
- Porciuncula et al., Wearable movement sensors for rehabilitation: A focused review of technological and clinical advances. PM R 10, 912–923 (2018).
- B. Marques et al., The use of wearable technology as an assessment tool to identify between-limb differences following ACL reconstruction: A scoping review. Phys. Ther. Sport 56, 15–27 (2022).
- Milosevic et al., Kinect and wearable inertial sensors for motor rehabilitation programs at home: State of the art and an experimental comparison. Biomed. Eng. Online 19, 25–39 (2020).
- Lee et al., IoT-enabled knee sleeve for home rehabilitation: A validation study. Sensors 24, 3345–3360 (2024).
Posted by buchanle on Monday, May 18, 2026 in May 2026.
Tags: IMU, Knee Rehabilitation, Motion Tracking, Wearable Sensors