Development of Novel Wearable Biosensor for Continuous Monitoring of Central Body Motion

Abstract

Accidental opioid overdose and Sudden Unexpected Death in Epilepsy (SUDEP) represent major forms of preventable mortality, often involving sudden-onset catastrophic events that could be survivable with rapid detection and intervention. The current physiological monitoring technologies are potentially applicable, but face challenges, including complex setups, poor patient compliance, high costs, and uncertainty about community-based use. Paradoxically, simple clinical observation in supervised injection facilities has proven highly effective, suggesting observable changes in central body motion may be sufficient to detect life-threatening events. We describe a novel wearable biosensor for continuous central body motion monitoring, offering a potential early warning system for life-threatening events. The biosensor incorporates a low-power, triaxial MEMS accelerometer within a discreet, chest-worn device, enabling long-term monitoring with minimal user burden. Two system architectures are described: stored data for retrospective analysis/research, and an in-development system for real-time overdose detection and response. Early user research highlights the importance of accuracy, discretion, and trust for adoption among people who use opioids. The initial clinical data collection, including the OD-SEEN study, demonstrates feasibility for capturing motion data during real-world opioid use. This technology represents a promising advancement in non-invasive monitoring, with potential to improve the outcomes for at-risk populations with multiple health conditions.

1. Introduction

1.1. The Problem: Life-Threatening Conditions Requiring Rapid Detection

Accidental drug overdose is the most common cause of accidental death in most developed countries, while SUDEP (Sudden Unexpected Death in Epilepsy) is the most common cause of death due to uncontrolled epilepsy [1,2]. Both of these conditions represent sudden catastrophic events, but if detected at onset and treatment is initiated rapidly, they are associated with increased chances of survival [3,4].

The scope of this challenge extends beyond these two conditions. Many other life-threatening situations are associated with sudden-onset catastrophic events, and often with unexpected death, including SIDS (Sudden Infant Death Syndrome), DIBS (Dead in Bed Syndrome), cardiac failure, and end-stage renal failure [5,6,7,8,9]. Though the exact mechanism of death in these conditions is not well studied, it is likely driven by the emergence of similar sudden-onset changes. While it remains unknown whether early detection and intervention would prevent fatalities, early detection could at least increase our understanding such that lifesaving responses can be developed [10,11,12].

The opioid overdose crisis exemplifies the urgency of this problem. In the United States alone, in 2023, over 105,000 people died from drug-involved overdoses, with opioids involved in the majority of these cases (72,776 involving synthetic opioids, primarily illicitly manufactured fentanyl) [13]. The pharmacology of opioid overdose primarily affects the central nervous system, with a sudden-onset fatal outcome [14,15,16].

1.2. Current Monitoring Limitations

The existing monitoring technologies face significant challenges that limit their effectiveness in preventing these deaths [17]. The current physiological monitoring methods (such as electrocardiography (ECG), electroencephalography (EEG), electromyography (EMG), end-tidal CO₂ monitoring, and polysomnography) can track vital signs in clinical settings, but typically require a complex setup, are limited to specific environments, and are consequently not suitable for continuous, long-term monitoring [17,18,19,20]. Additionally, these technologies often suffer from poor patient compliance, high cost and limited availability, limited battery life, and inconsistent data quality, particularly when home-based continuous monitoring is desired [17,19].

Paradoxically, the real-world evidence suggests that simple observational monitoring can be highly effective [17,21]. Globally, Medically Supervised Injecting Centres (MSICs) or Safe Drug Consumption Facilities, where people who use drugs are observed for the onset of overdose, have reported very low incidence of serious adverse outcomes, and virtually no fatalities over decades of operation [22,23,24,25,26,27]. Common to all is a centrally located observer—who may or may not be clinically trained—visually monitoring for the onset of overdose without the use of any vital sign monitoring devices [25,28]. This suggests that observable changes in central body motion may also provide crucial early indicators of life-threatening events [29].

1.3. Central Actigraphy as a Solution

This gap between the effectiveness of simple observation and the limitations of complex monitoring technologies points toward a novel solution: central actigraphy. Actigraphy, the continuous measurement of movement or activity, has been used in clinical and research settings for decades [30]. Traditionally, actigraphy devices have been worn on the wrist or ankle to monitor sleep–wake cycles and physical activity patterns [31]. However, the potential of using centrally worn actigraphy devices to monitor neurological status in real time for the detection of life-threatening events has been largely overlooked.

Central actigraphy presents a unique opportunity to bridge the gap between simple observational monitoring and complex physiological measurements, offering potential insights into neurological status through continuous monitoring of central body motion [21,32,33,34]. Motion sensors, particularly accelerometer and inertial measurement units (IMUs), have been widely used in consumer wearables for fitness tracking, fall detection in elderly populations, and sleep monitoring [35]. Recent studies have demonstrated the utility of accelerometer-based devices for detecting various physiological states, including seizure detection in patients with epilepsy [36,37]. Despite this technological maturity and established correlation between altered movement patterns and compromised neurological states, the application of motion sensors for overdose detection has remained largely unexplored until recently.

Several factors may explain this gap in application. First, the stigmatization of substance use has historically limited research investment in harm reduction technologies for people who use drugs [38]. Second, the heterogeneity of overdose presentations and individual drug use patterns creates more challenges than more predictable conditions, like falls and regular sleep–wake cycles [39]. Third, uncertainty about the temporal relationship between detectable motion changes and other physiological indicators has made it unclear whether monitoring could provide a meaningful early warning window for intervention.

While formal clinical diagnosis of opioid overdose is usually achieved through identification of respiratory depression and loss of consciousness, observers in drug consumption facilities typically rely, at least initially, on visual observation of body motion after drug ingestion. These motion changes are potentially detectable through centrally placed motion sensors before peripheral indicators such as oxygen saturation changes become apparent. A systematic review examining diagnostic definitions of opioid-induced respiratory depression (OIRD) relevant to accelerometry-based monitoring found that respiratory depression typically manifests with measurable changes in breathing patterns (including respiratory rates declining to 5 breaths per minute or below, and apneic episodes of 30 s or longer), which correspond to observable reductions in central body motion [40]. This approach could potentially provide early detection and alert for action in order to prevent deaths from accidental overdose, SUDEP, and potentially other conditions associated with sudden-onset disruptive changes which are nevertheless detectable.

By exploring the core capabilities of this technology, we aim to contribute to the growing field of wearable health monitoring devices and demonstrate how central actigraphy can provide a practical solution for research, and subsequently early detection of life-threatening events. This work represents a significant step forward in continuous physiological monitoring, with the potential to improve patient outcomes across a wide range of conditions associated with risk of early death. This paper describes the development and functionality of a device which can be configured to output linear acceleration measurements and/or rotational velocity measurements for different clinical applications.

The device described in this manuscript represents an extension of prior feasibility work demonstrating the technical capability of chest-worn accelerometry for monitoring drug use events [41]. This paper expands upon this initial work by describing the complete system architecture, presenting two distinct operational configurations (DCM and ALERT).

2. Device Development and Technical Overview

2.1. Hardware Design and Specifications

We have developed a small, wearable biosensor that incorporates a Micro-Electro-Mechanical System (MEMS) inertial motion unit capable of continuous central body motion monitoring over extended periods [41]. The device measures 40 mm in diameter and 14 mm in height, designed as a discreet, body-wearable biosensor, as seen in Figure 1. The primary objective is to provide a system that autonomously records movement, potentially recording early signs of life-threatening events across various conditions.

The biosensor incorporates a triaxial MEMS accelerometer, microcontroller unit (MCU) with Bluetooth capability, onboard memory for temporary data storage, and a lithium coin cell battery [CR2032]. The triaxial MEMS sensor provides high-resolution acceleration measurements [range ±2 g, 16-bit resolution] across three axes, enabling comprehensive capture of motion, as seen in Figure 2. The data acquisition rate is configurable from 12.5 Hz to 52 Hz. The measurement data is transmitted from the biosensor via Bluetooth for storage and/or downstream processing, depending on the use case and system architecture.

Several power optimization techniques are employed to extend battery life for extended monitoring periods; the system uses only linear acceleration output from the inertial measurement unit (IMU) (excluding rotational gyroscope measurements to reduce power consumption). This design choice provides substantial power savings, as demonstrated using typical current consumption values for the commercially available miniature 6-axis inertial motion units shown in

Table 1. Values shown are for high power/low noise modes. Data compiled from manufacturer datasheets.

Device	Accelerometer Only Current Consumption (mA)	Acc + Gyro Current Consumption (mA)
Bosch BMI330	0.15	0.79
ST LSMDSOX	0.17	0.55
TDK ICM-45686	0.12	0.42

. Additional power management strategies include minimizing Bluetooth data transmission through batch processing and limiting LED indicator usage.

The biosensor attaches using a standard ECG electrode fixation patch made of PET foam with acrylate skin adhesive, providing secure attachment, while maintaining skin compatibility during prolonged wear, as seen in Figure 1. The low-profile circular design of the device minimises interference with clothing and daily activities.

2.2. System Architecture Overview

We present two distinct configurations of this technology—the Data Collection and Management (DCM) system collects and stores data for retrospective analysis and research applications, and an in development Active Life-threatening Event Recognition Technology (ALERT) system for real-time collection of data plus processing, so as to offer detection and the potential of immediate intervention. Each system addresses different clinical and community needs, while sharing the core componentry and capability of continuous recording central body motion monitoring.

2.3. DCM System Architecture: Data Collection and Management

The Data Collection and Management (DCM) system implements a cloud-centric architecture that ensures reliable data capture and research applications. The biosensor transmits raw triaxial acceleration data via Bluetooth Low Energy (BLE) to the selected mobile application, which serves as a data relay with minimal on-device processing. The mobile application implements connection management, data buffering, and secure transmission protocols to ensure reliable data transfer to the cloud infrastructure, as shown in Figure 3.

The cloud infrastructure utilises scalable storage and processing capabilities designed to handle large datasets generated during continuous monitoring. The data are stored with full traceability and audit logging to meet potential future clinical research requirements. Figure 4 shows an example of the investigator dashboard, which utilising web-based access to collected data, provides advanced visualization tools, filtering capabilities, and export functions for statistical analysis software.

The system’s BLE implementation includes power management utilizing sleep states and batch transmission to minimise power consumption. Data transmission occurs in scheduled bursts rather than continuous streaming, extending the battery life, while maintaining data integrity through error detection and retransmission protocols.

3. ALERT System Architecture: Real-Time Detection and Response

The in-development ALERT system builds upon the DCM foundation, but utilises edge-computing architecture with local real-time processing capabilities, as illustrated in Figure 5. The same biosensor transmits the data to a specialised mobile application containing proprietary algorithms optimised for immediate sensor motion analysis. These algorithms run continuously on the mobile device, analysing the incoming sensor data streams for patterns indicative of reduced body motion.

The local processing approach eliminates dependency on internet connectivity for core functionality, enabling operation in environments with limited or intermittent network access. As seen in Figure 6, the mobile application includes sophisticated signal processing algorithms that filter movement artifacts associated with body motion patterns observed in real-world use situations.

Alert generation, as seen in Figure 7, involves multiple confirmation steps to minimise false alarms, while ensuring a rapid response to genuine emergencies. The system implements escalating alarm protocols; initial patient alerts through haptics, audio, and visual display notifications; and optional remote responder notifications if the patient fails to respond within specified timeframes.

4. Clinical Integration and Applications

4.1. DCM System: Application in Research and Clinical Practice

The DCM system is designed to address the identified need for comprehensive motion monitoring in research and clinical settings. Its continuous data capture capabilities enable collection of datasets that traditional episodic monitoring cannot provide, and the objective is to advance understanding of motion detection, including its relationship to the conscious level across various conditions.

In future clinical practice, the DCM system may contribute to detailed assessment of motion patterns in patients with suspected sleep disorders, neurological conditions, and those undergoing medication adjustments that might affect central motion in addition to more crisis-related detection of an overdose situation for example. The objective is that longitudinal data collection might enable tracking of disease progression and optimization of treatment plans.

4.2. ALERT System: Application for At-Risk Populations

The ALERT system is being developed to directly address the community monitoring gap which has been identified. For individuals at risk of opioid overdose, the system’s continuous monitoring has the potential to prevent deaths by providing the immediate detection and alerting that Supervised Injecting Facilities achieve through human observation.

The system’s real-time detection capabilities could perhaps also be used to enable early intervention with layperson naloxone administration and summoning of emergency services. The concept of a system of “Virtual” Supervised Drug Consumption is also currently being explored.

5. Clinical Testing and Real-World Evidence

User Acceptance and Acceptability

User acceptability is a crucial aspect of device testing as a part of the progress towards validation, particularly for technologies aimed at monitoring for potential opioid overdose amongst active drug users. A recent qualitative study identified the key factors influencing the adoption of wearable devices among people who use opioids (PWUO) [29]. The study found that device accuracy, ease of use, and discretion were critical to acceptability, with users emphasizing the importance of distinguishing between an overdose and a non-life-threatening state of opioid-induced sedation.

The acceptability and usability of the device continues to be evaluated through interviews and focus groups with volunteer participants. Analyses of the evidence gathered has identified three key thematic areas affecting device acceptance: device-specific factors (such as hardware comfort, software usability, and data security), person-specific factors (including technology access, drug-related differences, and individual circumstances), and environment-specific factors (like housing status and social networks) [29]. The key findings revealed that participants required devices that are inconspicuous, comfortable, accessible, easy to use, controlled by trustworthy organizations, and highly accurate. Accuracy of overdose detection was unanimously identified as the most important characteristic, with participants emphasizing the critical need for devices to differentiate between an overdose and an optimal state of being high. The barriers to adoption included affordability concerns, limited technology access, and trust issues, both in device accuracy and between technology providers and people who use opioids. The facilitators included inconspicuousness to avoid stigmatization, a longer battery life, and targeting specific subpopulations of people who use opioids. The study found that determinants of acceptability can be categorised into device, person, and environmental factors, with device developers needing to consider both the type of end user and their environment and how these influence acceptability [29].

This user feedback has informed the ongoing development of both the DCM and ALERT systems, with particular attention given to user experience factors that might influence acceptability and adherence in both clinical and community settings.

6. Ongoing Device Validation and Performance

6.1. The OD-SEEN Study: Real-World Clinical Data Gathering

An observational feasibility study, the OD-SEEN study, was undertaken at a Medically Supervised Injecting Centre (MSIC) in Sydney, Australia [42], to establish the feasibility of continuous motion monitoring during real-world drug use events in a supervised clinical environment and to collect valuable proof-of-concept data on the performance of central motion monitoring during real-world drug use events. This study captured motion data during 1145 actual heroin/opioid self-administration events, including 10 overdose events among 47 participants [41].

The OD-SEEN study utilised a functionally equivalent variant of the DCM system (same biosensor and system as shown in Figure 3, but with a Raspberry Pi replacing the tablet as the data transfer device).

The aim of the study was to capture chest motion data to establish technical feasibility rather than develop or validate detection algorithms enabling correlation between the motion sensor data and clinical observations of drug events. Analysis of data is ongoing to assess the extent to which the PneumoWave biosensor can detect the same motion pattern changes that trained observers watch for. The motion pattern examples presented in Figure 8 represent illustrative case studies selected to demonstrate the range of motion signatures observed across different event types. The results of these ongoing analyses will indicate the extent to which these device-derived quantitative data match, or can potentially enhance, the observational approach that has hitherto successfully prevented overdose deaths.

6.2. Preliminary Analysis of Motion Patterns During Drug Use Events

The threshold values referenced (motion values above 0.1 or below −0.1 indicating high activity, values between −0.1 and 0.1 suggesting sedentary movement, and values between −0.001 and 0.001 representing minimal movement) are descriptive classifications used for preliminary exploratory analysis rather than validated detection criteria.

These preliminary observational findings reveal distinct motion patterns across different types of drug events, demonstrating the potential for the PneumoWave biosensor to differentiate between safe use, concerning events, and an overdose. Figure 8 Illustrates three representative examples from the 1145 recorded injection events. During safe injection events (Figure 8A), the motion data consistently remained within the high-activity-to-sedentary range (values between −0.1 and 0.1), with minimal periods of low activity movement (values between −0.001 and 0.001). In contrast, concerning events that prompted staff intervention (Figure 8B) showed brief, but notable periods where motion fell into the low activity range immediately preceding staff prompts, suggesting detectable changes in movement patterns before clinical concern became apparent. Most significantly, confirmed overdose events (Figure 8C) demonstrated sustained periods of extremely low motion activity, with prolonged intervals where the readings remained within the minimal movement threshold. These preliminary findings suggest that central body motion monitoring can capture the same movement pattern changes that trained observers watch for, with quantifiable differences in motion signatures between safe use, concerning events, and a life-threatening overdose. The data suggests that the biosensor successfully detected the reduction in central body motion that characterised the onset of opioid-induced respiratory depression and loss of consciousness, supporting the hypothesis that wearable motion sensors can provide objective, quantitative measures to complement or enhance human observation in overdose detection.

7. Conclusions and Future Directions

The development of the PneumoWave biosensor is an important element of advancing the ability to monitor the central motion data of individuals at risk. Through procedures for technical validation, usability testing, clinical trials, and acceptability studies, continuous monitoring via an IMU-based sensor can be assessed to establish the reliability of detection of onset of lack of body motion with particular value for identifying potentially life-threatening conditions of low-level consciousness or low-level participant movement.

The validation process is currently ongoing and is measuring the device’s accuracy when compared to gold-standard bench-based methods, which will determine its usability for both healthcare professionals and end users and its acceptability among diverse populations, including people who use opioids. The findings broadly indicate that such technologies can address critical gaps in current monitoring approaches by providing continuous, non-invasive monitoring in both clinical and community settings.

Beyond technical validation, an important implementation issue will be the choice between the event-triggered monitoring approach and continuous 24/7 monitoring models. Each approach carries distinct implications for user acceptability, device utility, and clinical outcomes. The selection of an appropriate operational model will need to be guided by individual user values, concerns, and lifestyle circumstances rather than solely by technical capabilities or clinical preferences. As acceptability research has demonstrated, factors such as device discretion, user control, and alignment with personal risk perceptions significantly influence adoption and adherence. Future implementation strategies should therefore incorporate flexible approaches that allow users to choose between operational models and sensor thresholds based on their individual circumstances, risk profiles, and personal preferences, while ensuring that both the models maintain the accuracy and reliability essential for effective overdose detection and prevention.

Looking ahead, several important lines of investigation still need to be pursued to further advance this technology and its applications. Long-term effectiveness studies will be essential to examine the actual impact of this device (and others) on overdose outcomes in community settings and to determine real-world efficacy in preventing fatal overdoses. Continued development of algorithms to improve sensitivity and specificity for detecting different types of motion and disruption will require exploration and testing across diverse populations and in different, often challenging, environmental conditions—these are priority tasks.

By offering two distinct, but complementary approaches to monitoring—retrospective analysis through the DCM system for data capture for later analysis versus real-time detection through the ALERT system—the objective is to address the diverse needs of both clinical researchers and at-risk individuals, potentially transforming our approach to motion monitoring and improving outcomes across a range of conditions and challenging contexts.