Internet-of-Things Management of Medical Chairs and Wheelchairs

Chelsea Yeh

, Alexander W. Lee

, Hudson Kaleb Kalaw Dy

and Karin C. Li

Walnut Valley Research Institute, Walnut, California, U.S.A.

School of Medicine, University of California Riverside, Riverside, California, U.S.A.

Keywords: Wheelchair, Geri Chair, Treatment Chair, Hospital Management, Injury Prevention, Internet of Things, IoT.

Abstract: In this paper, we describe the application of the technologies of the Internet-of-Things (IoT) to the

management of wheelchairs and medical chairs such as geriatric (Geri) chairs or treatment chairs. Specifically,

it seeks to monitor the status of high-risk or physically-weakened patients in hospitals or care facilities as they

rest on wheelchairs or await treatment on medical chairs with sensor data collected by embedded pressure and

motion sensors, and provide real-time alerts to the medical staff. The potential for injuries from high-risk

individuals attempting to stand and falling is very serious. The injuries often result in additional complications

to the underlying health condition requiring the use of the wheelchair or treatment. The proposed IoT

wheelchair and medical chair management system will alert the staff immediately when a susceptible

individual stands or attempt to stand, and allow them to take immediate remedial action. The motion data

from the network of sensors is further processed by machine learning models which predict occupant intent

regarding sit-to-stand transition, providing preventive alerts to the staff. The research consists of two parts.

The first part created IoT-connected sensors and devices used to capture the occupant’s motion on the chair

and send the data to a central server. The second part developed the staff alert application that runs on mobile

phones and consoles located in the nurses' stations, that receive the information from the server.

1 INTRODUCTION

In a hospital or care facility, one of the roles of the

staff is to ensure the safety of patients as they rest or

wait for treatment. Internet of Things (IoT) allows for

real-time remote collection and interpretation of data,

and immediate feedback of monitored status based on

the collected data. In medical applications, IoT

enables the collection of medical data of patients in a

hospital and the dissemination of the patients’ status

to the medical staff immediately. This allows the

medical staff to assess the condition of those in their

care, and take appropriate actions to prevent or

mitigate worsening medical conditions or additional

complications and injuries.

In this research, we developed a real-time

wheelchair and medical chair monitoring system

based on IoT-connected pressure and motion sensors.

Here, wheelchairs and medical chairs are equipped

with IoT pressure and motion sensors that detect

whether or not the occupant has moved to a standing

position, or is attempting to stand. The devices

transmit the data of the motion or change in pressure

in real-time to a central server. Computational

intelligence based on the sensor data can be provided

on the chairs to detect and/or predict occupant status

or intent. The server processes and interprets the data,

then updates to a mobile application running on the

staff’s mobile devices, alerting them if a high-risk

occupant stands or is attempting to stand. It can also

alert the nurses in the nurses’ station via a dedicated

console, or be integrated into the facility's patient

management system.

Further work includes the measurement of the

patient’s lower body strength during recovery or

rehabilitation. Current tests include the 30-second

chair stand test (30CST) and 5 times sit-to-stand test

(5xSTS) are used to assess lower-body strength

(Zhang, 2018). Chairs equipped with pressure sensors

and motion sensors will be able to capture the

dynamic weight distribution and motion of the seat as

the patients perform the test, and provide more data

regarding the area of performance improvement and

strengthening.

2 SIT-TO-STAND EVENT

MANAGEMENT

The elderly population, patients who are physically

weakened due to medical conditions, and patients

whose medication impairs physical control are

Yeh, C., Lee, A., Dy, H. and Li, K.

Internet-of-Things Management of Medical Chairs and Wheelchairs.

DOI: 10.5220/0011059600003194

In Proceedings of the 7th International Conference on Internet of Things, Big Data and Security (IoTBDS 2022), pages 183-188

ISBN: 978-989-758-564-7; ISSN: 2184-4976

183

particularly susceptible to falls in injury. This is

especially problematic when individuals move from a

sitting position to a standing position (Pozaic, 2016).

In addition, patients with cognitive and memory

impairment may attempt to stand and leave while

resting in a sitting position, risking both falls and

becoming lost by wandering. Our project aims to

solve these issues by introducing a monitoring system

designed to supervise the movement of those

occupants while sitting in a wheelchair, geriatric

(Geri) chair, or other treatment chairs. This allows the

care staff the ability to react promptly and take

appropriate remedial actions to the high-risk

individual standing or attempting to stand. The idea

of putting Geri chair alarms to alert nurses and care

staff of occupant movement has been implemented

before in existing products (alzstore.com). In

previous works, chair alarms have been employed to

signal occupants leaving the chair with a co-located,

integrated alarm. These chair alarms consist of a

chord attached to an occupant’s clothing. If an

occupant stands, the cord pulls out a magnetic contact

sensor which produces an audible alarm, alerting the

caregivers of the situation. A pressure-sensitive pad

that rests on the seating area of the chair has also been

developed to sound an audible alarm if the occupant

stands, reducing the pressure on the pad.

However, in a hospital or care facility, this is

likely to disturb the other individuals in the vicinity

of the alarm. For our project, we aim to remove the

audio aspect of these alarms. By enabling the sensors

to send notifications to the staff via an IoT network

the disturbance created by the audio alarm is

eliminated. The staff must be alerted in case of an

alarm whenever they are on duty. Therefore, this

system sends alerts directly to the mobile phones of

the on-duty medical or care staff.

In our previous work, we integrated a pressure

sensor to hospital beds with IoT to alert nurses when

a bed-rest patient vacates their bed and risking falls

and injuries (Yeh, 2021). Another work incorporates

gesture recognition of data from a sensor array

mounted on patient chairs that provide an audio alarm

to the staff and PC notification (Knight, 2008).

3 COMPONENTS

The embedded IoT medical system consists of the

following elements:

1) Network of pressure sensors and motion sensors

suitable for wheelchair or medical chair use;

2) Internet-capable processing devices co-located

and connected to the chair absence sensors,

which collect, interpret, and transmit the sensor

data

3) An algorithm or set of algorithms, running on

the processing device and/or server, to

determine if the occupant has stood up or is in

the process of standing up. The algorithms can

also include machine learning models for

occupant intention-to-stand prediction

4) A WiFi, 5G, or other suitable and secure

network accessible from inside the facility or

environment

5) A data collection server that receives the chair

status data for the facility or the sector of the

facility which it is monitoring

6) Mobile devices or cellular phones that is

normally carried by the staff during their shift,

running the mobile application that displays the

status of the chairs that the staff is attending to

or responsible for

3.1 Sensors

Pressure sensors such as pressure pads are

commercially available and easy to obtain. These

devices sense pressure through the distribution of the

occupant weight among the pad and sensors. In this

system, the pressure pad is connected to a Wi-Fi

module. The purpose of the module is to read real-

time occupant data from the sensors and relay the

information to the data analysis server.

Also, motion sensors or accelerometers are

readily available to measure the movement on the

chair. Here, occupant motion can be monitored to

determine if a sit-to-stand event is occurring. In

addition, the accelerometer may be able to determine

the quality of rest that the occupant is experiencing.

We would like to emphasize that the accelerometer is

to be mounted on the chair, and not worn on the wrist

of the patient/user. This minimizes patient discomfort

and the workload on the care staff by reducing bodily

attachments, as the patient may already have various

monitoring devices and intravenous tubes attached to

their bodies.

In addition to pressure and motion sensors, it is

also possible to utilize other sensing technology to

assist in the monitoring of wheelchairs or medical

chairs. Heat sensors can also be deployed on the chair

to monitor the temperature of the chair as the

occupant is sitting on the chair. This temperature will

not be an accurate reading of the occupant's body

temperature, due to the lack of direct contact between

the sensor and the occupant's body (shielded by gown,

IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security

184

sheets, etc.). However, it does provide a relative

difference in temperature when the occupant is on the

chair versus not on the chair.

3.2 Microcontroller

The sensors will be connected via a suitable interface

to a computing device (microcontroller). Depending

on the computing and power requirements, the

microcontroller can also perform computation on the

collected data. For example, if the algorithm is simple

enough for the sensing modality, the microcontroller

would be able to directly determine if a sit-to-stand

transition has occurred.

The microcontroller must be equipped with

networking capabilities, preferably wireless, which

will be used to transmit the collected data to a server.

Ideally, the microcontroller will be powered via an

AC power adaptor. This provides a constant source of

power delivered by the facility. Since many treatment

chairs are typically already powered, using AC power

should be an option in a hospital environment.

However, for wheelchairs AC power is usually not

available because of their mobility, and rechargeable

batteries must be deployed. Even if AC power is

available, rechargeable batteries may be

advantageous to provide backup power to the

microcontroller and sensing devices in case of a

power failure.

3.3 Algorithm and Machine Learning

From the data collected by the sensors, one or more

algorithms are used to determine the status of the

occupant on the chair. Several types of algorithms can

be developed and deployed in the system, and can be

dynamically changed in response to clinical or patient

requirements. The proposed algorithm types are as

follows:

1) Simple chair vacancy algorithm

Determines whether the person has left the chair

2) Sit-to-stand transition algorithm

Determine in real-time what stage of the sit-to-

stand transition is being executed, from fully

seated, to weight shifting, to weight transfer, to

fully standing.

It will also determine if a sit-to-stand attempt

has failed.

3) Predictive sit-to-stand algorithm

Predicts from the movement on the sensors the

probability that the person is intending or begins

to attempt to stand

4) Lower body strength algorithm

Measures the dynamic movement of the patient

as they perform sit-to-stand transitions, and

determines rehabilitation progress and relative

strength and weakness during stages of

movement

With a pressure pad under the seating area, the

algorithm to determine if the occupant has vacated the

chair is relatively simple to implement a type 1

algorithm. Either a lower threshold pressure level has

been crossed, or a sudden decrease in pressure level

can be used to trigger an alert. With additional

pressure sensors, located on the four corners of the

seat, the weight distribution of the person can be

measured, allowing dynamic measurement of the

person’s motion to implement a type 2 algorithm.

This can be further enhanced by motion sensors

situated around the chair to implement type 3 or 4

algorithms.

Simple algorithms such as types 1 and 2 that do

not require much computing power can be

implemented directly in the microcontroller. This

would lower the data transmission on the

requirements on the network, and reduce the

computation load on the server. The microcontroller

can simply transmit the computed state of the person.

For more complex algorithms, if the

microcontroller lacks the necessary processing

power, it can simply transmit the received data to the

server for execution. However, with the capabilities

of modern microcontrollers, we expect more types of

algorithms to be executable locally on the chair.

We are planning the use of machine learning

models to predict the individual’s sit-to-stand

intention from the motion collected from the sensor

network. It is possible to use supervised learning to

train the model, with the researchers labeling the

intention of the seated individual prior to a sit-to-

stand event or no transition event. However, because

the time series data before an actual sit-to-stand event

is captured, this data can be used for unsupervised

learning, dramatically lowering the training cost of

the human labeling process. Although the training

will be done on the server, the pre-trained machine

learning model can potentially run on the

microcontroller.

3.4 Network

The hospital or care facility must provide a network

that the microcontrollers deployed on the chairs can

connect to. WiFi is the most prevalent wireless

networking system deployed in most modern

environments. The network must provide adequate

security against data interception and data breaches,

Internet-of-Things Management of Medical Chairs and Wheelchairs

185

and WiFi provides this capability. It must be highly

reliable (transmitted data is not lost or corrupted) and

have low latency (transmitted data is received

immediately). Bandwidth requirements are low for

the IoT chairs, as only a small amount of data is

transmitted per second per chair.

Other data networking systems may be used,

provided that the facility or environment possesses

such a system or is willing to invest in such a system,

and the system provides the necessary security.

Wireless systems include 5G cellular networks.

Wired systems that can be considered are ethernet,

powerline communication, coaxial, or telephone

networking.

3.5 Server

The role of the central server is to receive the data sent

by the microcontroller from sensors attached to the

IoT chairs. When a microcontroller or algorithm

running the server determines/predicts a sit-to-stand

event, it will send an alert to the mobile device or cell

phone carried by the responsible caretaker and/or the

console at the nurses’ station.

It is often the case that there are multiple nurses

or care staff in the facility, each attending to a set of

chairs, some with high-risk or physically-weakened

individuals, and some without.

The configuration of the assignment of the chairs

to each caretaker must be simple and straightforward.

Also, the chair alert must be able to be turned on or

off for each chair depending on whether a high-risk

individual has been placed on the chair, and if the

occupant has been removed from the chair (by the

staff) for reasons such as return to bed or treatment

sessions.

3.6 Mobile Devices and Mobile

Application

The mobile application that notifies and alerts the

responsible nurses or care staff of sit-to-stand

transition will run on a mobile device that the staff

member carries. For caretakers with facility-issued or

personal mobile phones that they carry on their

persons during their shift, the mobile application can

be installed and provisioned on these devices by the

IT staff. Dedicated devices can also be developed that

connect to the facility’s network and alert the

caretakers of risky sit-to-stand events.

As a high-risk or physically-weakened individual

is placed on a particular IoT chair, the nurse will be

able to scan a barcode or QR code, or connect via

Bluetooth or RFID, and turn on the alert for that chair.

Similarly, the alert is turned off for the chair prior to

when the occupant is removed from the chair.

The facility management or IT staff will be able

to assign the set of chairs to each particular mobile

device/caretaker from the central server, or on the

mobile devices.

The mobile device can provide alerts in the form

of (1) visible notification on the screen, (2) vibration

of the device, and/or (3) one-time or periodic audible

beep or alarm.

4 INITIAL SYSTEM

For the initial presence sensing system development,

we created a test system from the following

components.

The microcontroller was chosen to be the ESP32

because it is powerful, low cost, and, most

importantly, has integrated WiFi capabilities

(espressive.com). It is compatible with the Arduino

Integrated Development Environment, a widely used

open-source development system that uses the C++

coding language, making it simple for programming

and updating. The ESP32 platform features a built-in

dual-core CPU with WiFi connectivity, a wide

operating temperature range from -40°C to 125°C.

This development board is low-cost and usually

operates at 160MHz with 4MB of flash memory and

8MB of PSRAM.

WiFi networks are ubiquitous and secured with

WPA2 (IEEE, 2016). The IoT devices will be

connected over a 2.4GHz WiFi band because it is

more secure and has a longer range compared to the

5.0GHz band.

For this initial development, we chose the Blynk

cloud-based server. Blynk is an IoT platform that

allows machine learning and data analytics on mobile

apps and runs over the HTTPS API. Blynk (blynk.io)

operates with the concept of a virtual pin, which

enables data to be exchanged from a device to the

server easily via the cloud.

Similarly, we used the commercially available

Blynk application for the application that runs on

mobile devices. It operates on both iOS and Android.

The Blynk mobile application uses a drag-and-drop

interface, allowing for a simple and user-friendly

design.

We are investigating and building the different

types of algorithms that will process the sensor data.

We are building the machine learning models using

TensorFlow (Abadi, 2016), and are investigating the

use of TensorFlow Lite Micro (David, 2020) to run

the pre-training model on the ESP32.

IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security

186

For future development, we plan to migrate to

more secure and robust servers and mobile devices.

Cloud-based servers include Amazon Web Services

(Amazon 2015), Google Cloud (Gonzales, 2015), and

Microsoft Azure (Copeland, 2015). Also, a variety of

software solutions exist if the facility plans to host the

servers in-house.

The React Native environment is being planned

for the development of the mobile application. It

supports both the Android and iOS operating systems.

React Native (facebook.github.io) is built on

JavaScript while rendering in the platform’s native

user interface, enabling cross-platform sharing of

code from a single codebase.

Figure 1 shows the screenshot of the chair

management mobile application on the iOS

environment. Here, alerts for Chair 1 and Chair 4 are

active. Chair 2 and Chair 3 are turned off because no

high-risk individuals are occupying the chairs. Chair

1 is showing green, meaning the occupant is in the

chair. Chair 4 is showing red, meaning the occupant

has stood up and an alarm is raised.

Figure 1: Alarms; Chair alert on/off.

Figure 2 shows the screenshot of the mobile

notification alert in real-time as the occupant on Chair

4 stands up.

Figure 3 shows the screenshot of the alert on the

lock screen of the iOS device in real-time as the

occupant on Chair 4 stands up.

Figure 2: Notification on the application screen.

Figure 3: Notification on the lock screen.

Figure 4 shows a commercially available pressure

sensing pad (nationalcallsystems.com) used to

monitor occupant presence on the chair.

Internet-of-Things Management of Medical Chairs and Wheelchairs

187

Figure 5 shows a semiconductor motion sensor

(InvenSense MPU-6050 accelerometer + gyroscope)

(InvenSense, 2020) module and a strain gauge weight

sensor (sparkfun.com).

Figure 4: Pressure sensing pad.

Figure 5: Motion and weight sensor modules.

5 CONCLUSION AND FUTURE

WORK

We have created a novel IoT-based wheelchair and

medical chair management system that uses

integrated sensors (pressure and motion sensors),

WIFI-connected microcontrollers, algorithms, cloud-

based servers, and mobile applications. This alerts the

nurses and care staff immediately of sit-to-stand

events that may lead to falls and serious injuries, and

take mitigating actions to prevent such injuries and

complications.

We propose the integration of additional sensors

and algorithms to the medical chairs that enable the

dynamic assessment of patient rehabilitation and

recovery in each phase of the sit-to-stand transition,

and identify areas of strengthening and weakening.

REFERENCES

Zhang, Q., Li, Y. X., Li, X. L., Yin, Y., Li, R. L., Qiao, X.,

& Hou, G. (2018). A comparative study of the five-

repetition sit-to-stand test and the 30-second sit-to-

stand test to assess exercise tolerance in COPD

patients. International journal of chronic obstructive

pulmonary disease, 13, 2833.

Pozaic, T., Lindemann, U., Grebe, A. K., & Stork, W.

(2016). Sit-to-stand transition reveals acute fall risk in

activities of daily living. IEEE journal of translational

engineering in health and medicine, 4, 1-11.

https://www.alzstore.com/fall-prevention-alarm-

p/0064.htm

Yeh, K., Yeh, C., & Li, K. (2021). Internet-of-Things

Management of Hospital Beds for Bed-Rest Patients.

In Advances in Computer Vision and Computational

Biology (pp. 439-448). Springer, Cham.

Knight, H., Lee, J. K., & Ma, H. (2008, August). Chair

Alarm for patient fall prevention based on Gesture

Recognition and Interactivity. In 2008 30th Annual

International Conference of the IEEE Engineering in

Medicine and Biology Society (pp. 3698-3701). IEEE.

https://www.espressif.com/en/products/hardware/esp32/

overview

IEEE Standard for Information technology—

Telecommunications and information exchange

between systems Local and metropolitan area

networks—Specific requirements - Part 11: Wireless

LAN Medium Access Control (MAC) and Physical

Layer (PHY) Specifications," in IEEE Std 802.11-2016

(Revision of IEEE Std 802.11-2012), vol., no., pp.1-

3534, 14 Dec. 2016

https://blynk.io/

Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean,

J., & Zheng, X. (2016). Tensorflow: A system for large-

scale machine learning. In 12th {USENIX} symposium

on operating systems design and implementation

({OSDI} 16) (pp. 265-283).

David, R., Duke, J., Jain, A., Reddi, V. J., Jeffries, N., Li,

J., & Warden, P. (2020). Tensorflow lite micro:

Embedded machine learning on tinyml systems. arXiv

preprint arXiv:2010.08678.

Amazon, E. C. (2015). Amazon web services. Available in:

http://aws. amazon. com/es/ec2/(November 2012), 39.

Gonzales, J. U., & Krishnan, S. P. T. (2015). Building your

next big thing with Google Cloud Platform. Aprés, 27.

Copeland, M., Soh, J., Puca, A., Manning, M., & Gollob,

D. (2015). Microsoft Azure. New York, NY, USA:

Apress.

https://facebook.github.io/react-native/

https://www.nationalcallsystems.com/chair-sensor-pad-7-

x-15-45-day

InvenSense, T. D. K., & CT, P. C. T. (2020). MPU-

6050. TDX Invensense.

https://www.sparkfun.com/products/10245

IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security

188