Move towards smarter clothing by creating a compression sleeve that continuously monitors its pressure and how often it’s being worn

video: overview | video: conference presentation | making sensors | experiments | publications

Collaborators: Cheryl Brunelle, Alphonse Taghian, Daniela Rus

As the world of small wearables enters the world of big data, new possibilities arise for smart clothing that can improve quality of life – basically supersuits that can sense our activity and provide physical and cognitive support. One exciting use for these will be medical applications. Instead of evaluating progress at discrete times whenever a patient visits a doctor’s office, wearables can provide continuous streams of personalized data to reveal how metrics and garments evolve over time. These tools and datasets could allow doctors to provide a higher level of care – they can learn more about a patient’s unique circumstances, and do better research about what prescriptions are most effective for the general population.

Taking a step towards these smarter wearable devices that can improve our diagnoses and treatments, this project presents a smart compression sleeve with integrated soft pressure sensors. It aims to help improve the treatment of lymphedema, which is a potentially irreversible swelling of the arm that may occur after breast cancer surgery. The state of the art for treatment is to prescribe wearing compression sleeves that exert pressures of 20-30 mmHg for at least 12 hours per day, but it is hard to know whether this is the most effective prescription and whether patients adhere to it.

To address this data gap and to provide doctors with enhanced tools, this project presents and evaluates soft pressure sensors that are unobtrusively integrated into a compression sleeve to continuously monitor its pressure and how long it is worn.

We present a soft pouch-based pneumatic sensor and its application to a wearable monitoring system, and compare it to a commercially available resistive sensor. Experiments investigate accuracy, characterization, calibration, and robustness to ambient conditions. Results can benefit deployments of soft pressure sensing for wearable devices or robotic applications.

Making the sensors

There are a few key metrics to consider when designing and fabricating sensors for the compression sleeve:

• Accuracy and repeatability: should measure exerted pressure and usage duration reliably enough for lymphedema treatment evaluation. Based on consultations with expert clinicians, pressure should be measured to within approximately 2 mmHg and usage duration should be estimated to within approximately 15 minutes over a 12-hour period.
• Varying environmental conditions: performance should be reliable despite changes in conditions including ambient temperature or pressure, such as traveling to cities with different climates and elevations.
• Soft and dynamic interfaces: the human arm and the stretchable sleeve are both deformable, creating challenging conditions for consistent pressure measurement.
• Daily wear: the sensors should be robust to common activities
• Comfort and safety: sensors should be unobtrusive, such that they do not cause the user any discomfort or adverse effects beyond what the sleeve alone would cause.
• Scalability and customizability: the fabrication approach should streamline sleeve integration and future widespread deployments.

Making the pneumatic pouch sensors

The pouch sensor consists of an air bladder connected to a pressure transducer. It is fabricated using an index or middle finger of a large vinyl powder-free medical examination glove from MedPride. Flexible silicone tubing with a 1.5mm inner diameter, a 3mm outer diameter, and a durometer of 50A is inserted approximately 1cm into the pouch. The interface is sealed using Smooth-On Sil-Poxy.

The current implementation uses a D2-P4V-Mini pressure transducer from All Sensors. Its operating pressure range is +/-30 inches of water (+/- 56 mmHg), and it outputs an analog voltage proportional to the difference in pressure between two input ports. The pouch tubing is connected to the positive port, and the negative port can be left open or sealed.

This approach uses a simple fabrication method and commodity materials to yield a soft robust pressure sensor with linear response. Its softness is important for user comfort and safety, especially in high-pressure and long-duration applications. It can also be rapidly prototyped and customized. Key design considerations include the pouch size, how much it is inflated, and whether to use an open or closed reference port.

Preparing the resistive sensors

A commercially available force-sensitive resistor (FSR) was used to explore resistive pressure sensing, namely the circular FSR 402 from Interlink Electronics. It is 0.6 mm thick, has a 1.8 cm diameter, and its active area has a 1.3 cm diameter. Its nominal sensitivity range is 0.1-10.0 N (5.7 – 565 mmHg if force is distributed over the entire active area). Connecting the sensor in a voltage divider yields a voltage that depends on applied pressure.

The sensor’s thin profile and small surface area allow it to be unobtrusive, but also allow it to sink into indentations of the skin and yield inaccurate results. Furthermore, uniformly stretching the sleeve over the sensor may causes its inactive edge to support much of the pressure and preclude successful measurements. To address these challenges, the sensor is fastened within a custom plastic sheath. White flexible plastic polystyrene sheets with a thickness of 0.5 mm were used. The FSR is glued to a 2.5 cm square to avoid sinking into the arm. A circle with a diameter of 1.1 cm is then centered over the active area, lightly taped on one side to prevent movement while not applying pressure. Finally, a 2.0 cm square is lightly taped on top. This sheath provides a larger surface area for the sleeve interface, and focuses exerted pressure onto the sensor’s active area.

This approach uses commercially available sensors and materials to create a thin, unobtrusive sensing layer. However, the need for a semi-rigid plastic sheath introduces edges on the skin that may be undesirable for long-term usage.

List of parts and materials

Testing how sensors respond to ambient conditions

An important goal for the deployable wearable system is to maintain medically relevant accuracy while reducing how much calibration is required. This is coupled with reducing sensitivity to ambient conditions; for example, if the sleeve is fitted in an air-conditioned doctor’s office at sea level, and then the patient travels to an elevated city on a hot day, the sensor would ideally continue to generate actionable data without requiring the user to calibrate. This goal can also help inform sensor design parameters.

Sensors were placed in an oven and a vacuum chamber to simulate varying ambient temperatures and pressures. Mock arms were created by wrapping medium-density foam around metal cylinders. Pneumatic or resistive sensors were placed around each cylinder, and Juzo compression sleeves were stretched on top. The five pouch sensors were fabricated with varying sizes to study how size impacts sensitivity.

Results suggest that larger pouches are less sensitive to changes in ambient temperature and pressure. A possible reason for this can be derived by considering the ideal gas law and the properties of elastic materials (see the publication for more information). Results also suggest that using an open reference is less sensitive than a closed reference, which can be considered using similar reasoning.

Based on these experiments, larger pouches with an open reference port are desirable to minimize ambient sensitivity. Balancing this with the desire to reduce obtrusiveness by minimizing size, a medium pouch with an open reference port was selected. The results for this size are promising for not requiring users to recalibrate as ambient conditions vary.

Most of the FSRs exhibit comparable temperature sensitivity as the pouch sensors within typical weather conditions, but are significantly more sensitive to higher temperatures. This may be due in part to the properties of the resistive material, and to the nonlinear relationship between resistance and pressure.

Characterizing the sensors

Ideally, the sleeve will be plug-and-play: a new user could don a new sleeve and immediately begin receiving personalized treatment feedback. This means that it’s important to know how accurate each sensor type can be when placed under a compression sleeve, and how much calibration will be needed. To address these aspects, the response curve of each sensor is characterized to convert voltage readings to pressure values and infer the expected accuracy during real-world deployments.

A testing rig was constructed to systematically collect data that maps voltages to ground-truth pressures. Key considerations are to be as close to real-world conditions as possible, including how pressure is exerted and the deformability of materials, and to enable systematic pressure variation. The presented setup aims to mimic a variable-diameter arm wearing a compression sleeve. A sensor being evaluated is lightly taped to one half, and a Juzo Pressure Monitor is inserted into the other half; the latter is used to measure sleeves during clinical visits, and can serve as the ground-truth measurement. As the rig is spread, the sleeve stretches and exerts increasing pressure on both the sensor and the Juzo monitor.

Data is streamed and synchronized from sensors and a webcam using the ActionSense multimodal recording framework. Given the target pressure ranges of compression therapy, the rig was spread to exert pressures between 10 and 50 mmHg at intervals of 2 mmHg.

Parameterized curves were derived that can explain the data from each sensor type, and then they were fit to the data from various combinations of sensors. To test accuracy without any calibration, curves were fit to data from a single sensor and then used to predict pressures from another sensor; these results are the top confusion matrix in the above plots. To improve the accuracy while still avoiding calibration, curves were fit to aggregated data from all sensors except one and then using that curve to predict pressures from the left-out sensor.

To increase accuracy further, a minimal calibration procedure was also investigated that assumes a patient has their sleeve measured once at a clinical visit. The current analysis implements this paradigm by fitting a curve to all sensors except one, then adjusting the curve for the left-out sensor based on a single calibration point.

The results suggest that the pneumatic pouch sensors would be sufficiently accurate for the clinical purposes even without any calibration, and become even more accurate with the single-point calibration method. The resistive sensors, however, have more variability between sensors and therefore would likely require a calibration procedure. With the single-point calibration, their accuracy becomes comparable to the pneumatic sensors.

Deploying the smart compression sleeves

Deployable smart sleeves were fabricated and worn for extended periods of time to demonstrate the feasibility of using the sensors for continuously monitoring pressure and usage duration.

Patches of thin fabric were affixed to the inside of compression sleeves at the forearm and upper arm, to create sensor pockets. These hold sensors in consistent positions, prevent sensors from contacting the user’s arm directly, and allow sensors to be removable.

A small wearable control board was created based around the Adafruit QT Py ESP32-C3 microcontroller, which features WiFi and Bluetooth. For improved accuracy, an ADS1115 ADC measures the sensor outputs and regulated voltage. A 500 mAh single-cell lithium-ion battery powers the circuits. These components, along with a charging circuit and FSR voltage dividers, are mounted on a 3 cm x 4 cm breadboard.

The microcontroller samples each sensor every second. Every minute, it computes the mean and standard deviation of the raw values. It then uploads these statistics to Google Sheets via WiFi, where a real-time dashboard provides pressure and usage information.

Two experiments were conducted. The first experiment used a single Class~I size~II sleeve worn on the left arm. It had a resistive sensor on the middle of the outer forearm, a pouch sensor on the outer forearm near the wrist, and a pouch sensor on the outer upper arm. It was worn for 71 hours over 7 days. The second experiment used two Class~II sleeves, each with a sensor on the outer forearm near the wrist and on the outer upper arm. A size~II sleeve worn on the left arm had pouch sensors, and a size~I sleeve worn on the right arm had resistive sensors. These were worn for 45 hours over 3 days. A single able-bodied right-handed subject participated in these preliminary tests of feasibility.

The sleeve systems successfully demonstrated continuous monitoring of both pressure and usage. The sensors were qualitatively unobtrusive, not causing noticeable discomfort beyond what the sleeve alone induces. The sensors produced reliable data throughout a variety of activities of daily living including working at desks, cooking, eating, walking, and commuting a few miles per day using a kick scooter. The experiments also spanned multiple ambient indoor and outdoor temperatures. The battery lasted a full day without recharging; this lifetime could also be substantially extended by reducing the frequency with which WiFi is used and by using low-power microcontroller modes.

Using the Juzo monitor to acquire intermittent ground truth estimates, results suggest that the pouch sensors generally achieved the accuracy goals for evaluating compression therapy and that they yielded lower errors than the resistive sensors. Both sensors successfully reflect a pressure gradient between the forearm and upper arm that the sleeves are designed to exert.

Based on the pressure measurements, a threshold can be used to determine whether the sleeve is being worn. During these initial tests, the recommended dosage was achieved on 60% of days. The current system is capable of measuring adherence with approximately 1-minute resolution.

Discussion

Both sensing modalities explored in this work demonstrate promise for continuously monitoring pressure and usage duration. They each have trade-offs regarding accuracy, integration within the sleeve, and robustness.

The pouch sensors exhibit a linear response, yield accurate results even without calibration, and can be rapidly fabricated and customized using commodity materials. Importantly, they are also soft and thus well-suited to wearable applications. However, the transducer or tubing must be mounted on the sleeve which may slightly increase obtrusiveness, and the transducer requires an outlet to the atmosphere which may complicate encapsulation for protection and waterproofing. Popping and leakages are also concerns.

The resistive sensors can yield accurate results, do not require a separate transducer mounted on the sleeve, and are generally physically robust. However, they require semi-rigid sheathing to appropriately distribute pressure onto the sensor, which introduces edges on the person’s skin under high pressures. They also exhibit higher sensitivity to ambient temperature, and characterization results suggest that at least a single-point calibration routine is required to obtain sufficient accuracy. The response curves can also be highly sensitive to where pressure is exerted on the sensor, so may change over time due to the sheath moving or to the sensor properties.

All together, the trade-offs and experimental results indicate that the pouch sensors are promising for wearable applications including compression therapy assessment. They exhibit comparable or improved performance compared to resistive sensors, minimizing calibration routines while achieving medically relevant pressure accuracies and usage duration resolutions. They are also soft and thus safer for prolonged human contact. Further work can improve robustness, and miniaturize and protect the transducer.

Future work will include longer deployments with a larger subject pool and close collaboration with clinical staff. Additional characterizations can also be performed with more instances of each sensor, and additional design parameters such as materials can be explored. Future work can also improve fabrication and integration, such as sewing conductive threads into the sleeve and extending battery life.

This work takes a step towards new insights for lymphedema treatment, and towards smart wearables that provide doctors and patients with real-time personalized feedback.

Conference Presentation: IEEE International Conference on Soft Robotics (RoboSoft 2023)

Publications

• J. DelPreto, C. L. Brunelle, A. G. Taghian, and D. Rus, “Sensorizing a Compression Sleeve for Continuous Pressure Monitoring and Lymphedema Treatment Using Pneumatic or Resistive Sensors,” IEEE International Conference on Soft Robotics (RoboSoft), 2023.
  [BibTeX] [Abstract] [Download PDF]
  

  Smart soft wearable devices have great potential to change how technology is integrated into daily life. A particularly impactful and growing application is continuous medical monitoring; being able to stream physiological and behavioral information creates personalized datasets that can lead to more tailored treatments, diagnoses, and research. An area that can greatly benefit from these developments is lymphedema management, which aims to prevent a potentially irreversible swelling of limbs due to causes such as breast cancer surgeries. Compression sleeves are the state of the art for treatment, but many open questions remain regarding effective pressure and usage prescriptions. To help address these, this work presents a soft pressure sensor, a way to integrate it into wearable devices, and sensorized compression sleeves that continuously monitor pressure and usage. There are significant challenges to developing sensors for high-pressure applications on the human body, including operating between soft compliant interfaces, being safe and unobtrusive, and reducing calibration for new users. This work compares two sensing approaches for wearable applications: a custom pouch-based pneumatic sensor, and a commercially available resistive sensor. Experiments systematically explore design considerations including sensitivity to ambient temperature and pressure, characterize sensor response curves, and evaluate expected accuracies and required calibrations. Sensors are then integrated into compression sleeves and worn for over 115 hours spanning 10 days.

  @article{delpreto2023sensorizedCompressionSleeve,
  title={Sensorizing a Compression Sleeve for Continuous Pressure Monitoring and Lymphedema Treatment Using Pneumatic or Resistive Sensors},
  author={DelPreto, Joseph and Brunelle, Cheryl L. and Taghian, Alphonse G. and Rus, Daniela},
  journal={IEEE International Conference on Soft Robotics (RoboSoft)},
  organization={IEEE},
  year={2023},
  month={April},
  url={http://www.josephdelpreto.com/wp-content/uploads/2023/04/DelPreto_smart-sleeve_RoboSoft2023.pdf},
  abstract={Smart soft wearable devices have great potential to change how technology is integrated into daily life. A particularly impactful and growing application is continuous medical monitoring; being able to stream physiological and behavioral information creates personalized datasets that can lead to more tailored treatments, diagnoses, and research. An area that can greatly benefit from these developments is lymphedema management, which aims to prevent a potentially irreversible swelling of limbs due to causes such as breast cancer surgeries. Compression sleeves are the state of the art for treatment, but many open questions remain regarding effective pressure and usage prescriptions. To help address these, this work presents a soft pressure sensor, a way to integrate it into wearable devices, and sensorized compression sleeves that continuously monitor pressure and usage. There are significant challenges to developing sensors for high-pressure applications on the human body, including operating between soft compliant interfaces, being safe and unobtrusive, and reducing calibration for new users. This work compares two sensing approaches for wearable applications: a custom pouch-based pneumatic sensor, and a commercially available resistive sensor. Experiments systematically explore design considerations including sensitivity to ambient temperature and pressure, characterize sensor response curves, and evaluate expected accuracies and required calibrations. Sensors are then integrated into compression sleeves and worn for over 115 hours spanning 10 days.}
  }