Sensors and Sensing Methods for Early Detection of Life-Threatening Sudden Illnesses in Motor Vehicles Drivers ^†

Abstract

Due to the increasing number of vehicles and the aging population, the vulnerability to sudden medical emergencies among drivers is a growing problem. Events such as heart attack, stroke, and loss of consciousness can occur without warning and endanger everyone on the road. Modern vehicles, equipped with electronic systems, can support real-time driver’s health monitoring through early detection technologies. The existing Driver Monitoring Systems (DMS) in our cars assess behavioral states such as drowsiness and distraction. In the future, DMS will include biometric sensors to monitor vital signs such as heart rate and respiration. By finding predictors of sudden illnesses (SI), such a system will provide valuable time for the driver to react before the strike of a medical event. In this paper, we present our vision for DMS operation with physiological monitoring capabilities. A brief overview of sensor’s types and their locations in the vehicle interior used in the research studies for monitoring the corresponding physiological parameters is presented. A comparative analysis of the advantages and disadvantages of the sensing methods used for physiological monitoring of the driver in real driving scenarios is made.

1. Introduction

Road accidents are a global issue. The increasing number of vehicles in everyday life is increasing the probability of fatal collisions. Meanwhile, the population of the developed countries is getting older and more concentrated in urban areas. As a result, the drivers are vulnerable to sudden medical emergency events. When the two factors are combined, a growing number of road accidents is expected in the future due to driver’s heart attacks, strokes, or loss of consciousness. These events can strike without warning, putting other drivers, passengers, pedestrians, and other road users in danger.

Finding a solution to this problem could significantly reduce potential tragedies and prevent serious consequences. Since our modern vehicles are equipped with advanced electronic systems, they can offer a technological solution to the problem by continuous monitoring the driver’s health status in real time through early detection methods.

According to the World Health Organization’s annual report from 2023 [1], fatal accidents on a global scale are decreasing. In the European Union, road fatalities are also decreasing. One of the reasons is the introduction of safety regulations such as the General Safety Regulation II (GSR II), which requires automotive manufacturers to implement monitoring of the driver by using a DMS.

In modern automobiles, this advanced system can determine the driver’s conditions, including awareness, distraction, drowsiness, and fatigue. Determination of eye gaze, driver’s posture, behavior, and use of alcohol and drugs are also features of DMS. Consequences of potential accidents related to the above factors are mitigated by this system. Its safety contribution may be expanded if the DMS features include health-related monitoring. This approach aims to address the problem of drivers experiencing life-threatening SI.

1.1. Statistics of SI Events

In 2008, Trevo et al. [2] concluded from statistics for the period 2003–2004, involving 522 deceased drivers, that 10.3% (N = 54) of the cases were due to sudden medical emergencies, with 70% (N = 38) of those caused by cardiac problems.

Also in 2008, Lindsay [3] systematized data on crashes where sudden medical conditions in drivers were confirmed as the cause and an ambulance was called. These data cover the period from April 2002 to October 2005 in Adelaide, Australia. In 19 (76%) of the 25 cases, the driver reported complete or partial loss of consciousness leading to the crash. In most of the cases (28%), drivers lost consciousness due to a sudden cardiac event.

In 2018, Brodie et al. [4] reported that approximately 10% of driver fatalities in Victoria, Australia, during 2012–2013 were due to SI. These events occurred either suddenly or with preceding symptoms within one hour. The most common causes were cerebral or cardiovascular diseases.

In 2020, Miao et al. [5] investigated 21 driver fatalities due to sudden ischemic heart disease during the period 2015–2019. Except for two, the remaining 19 drivers had no previous complaints or noticeable symptoms before the SI occurred.

In 2021, the Finnish Crash Data Institute (OTI) [6] published statistics for the period 2014–2018, indicating that 16% of 907 fatal incidents were due to a SI of the driver. These incidents involved 179 individuals, 142 of whom died. In 88% of the cases, the driver was male with a median age of 66, and 84.39% of the cases were related to cardiovascular events.

In 2023, Skyving et al. [7] published findings from a study conducted between 2010–2019 on 762 driver fatalities in Sweden involving individuals over 50 years of age. Of these, 29% were caused by SI, with the main causes being deteriorated heart conditions such as arrhythmias, strokes, and heart attacks. A smaller percentage of cases included hypo/hyperglycemia, fainting, and epileptic seizures.

Table 1. Published statistical data about fatal road accidents caused by SI after the year 2000.

Period	Country	Sudden Illnesses Share	Causes of Sudden Illnesses	References
2003–2004	Finland	10% of 522	70% cardiovascular diseases	Trevo et al. [2]
2002–2005	Australia	25 cases	28% cardiovascular diseases	Lindsay [3]
2012–2013	Australia	10%	Cardiovascular and cerebral diseases	Brodie et al. [4]
2015–2019	China	19 cases	100% sudden ischemic heart diseases	Miao et al. [5]
2014–2018	Finland	16% of 907	84.39% cardiovascular diseases	OTI [6]
2010–2019	Sweden	29% of 762	Cardiovascular and cerebral diseases	Skyving et al. [7]

summarizes the above publication’s results. It can be concluded that most cases of SI events are caused by heart problems since the most valuable physiological parameter is Heart Rate (HR).

1.2. Predictors of SI in Physiological Signals and Parameters

Due to the sudden nature of the medical events on the road, there is no dedicated, detailed database with abnormal physiological parameters or signals data that occurred during accidents. Such data may be used in DMS for training and tracking to identify distorted physiological signals or abnormal physiological parameters. Logical sources of reliable physiological data and knowledge are hospital databases and emergency rooms. The most common patients in the emergency rooms are those with cardiovascular and respiratory diseases. Clinical experience and research show that the changes of certain signal parameters and features can predict medical events before the condition or symptoms develop rapidly. Those signal changes are predictors of SI.

Chowdhury et al. [8] describe the morphology of electrocardiogram (ECG) signals during the onset of three types of myocardial infarction as examples of acute cardiovascular events. These ECG patterns are well-known ST-segment depression, inversion, and elevation. Such changes in the morphology of the ST-segment and T-wave are established markers of myocardial ischemia and infarction, and can be sought when analyzing real-time ECG recordings from vehicle drivers.

Other clinically relevant data applicable to DMS include the identification of predictors for acute cardiovascular events. For example, in 2015, Andersen et al. [9] reported that when monitoring patients in clinical settings who had experienced cardiac arrest, they were often observed to show signs of physiological deterioration prior to the event. The aim of this study was to determine the trend in changes of vital parameters from 1 to 4 h before cardiac arrest. The results indicate that, within this time frame prior to the medical event, at least one vital parameter was outside normal limits in approximately 60% of the patients.

In 2023, Sasaki et al. [10] concluded that a prolonged QRS complex duration (>120ms) on the ECG at hospital admission is a strong predictor of in-hospital mortality among patients with acute myocardial infarction.

Based on these findings, it can be concluded that future DMS aiming to detect acute cardiovascular events should monitor the driver’s health status continuously and persistently. This assessment can be performed by capturing physiological signals from the driver’s body and monitoring certain physiological parameters, which provide essential feedback about the driver’s condition and serve as indicators of emerging uncontrolled medical issues. In such potentially fatal situations, it is crucial for the driver to stop the vehicle to ensure his/her own safety, as well as that of any passengers and other nearby road users.

1.3. Physiological Parameters and Signals

The important physiological parameters (vital signs) for monitoring are: Heart Rate (HR), Heart Rate Variability (HRV), Respiratory Rate (RR), Blood Pressure (BP), Body Temperature (BT), Oxygen Saturation (SpO2), and Electrodermal Activity (EDA) [11]. To determine their values, the physiological signals must be obtained from the driver’s body through well-known techniques such as ECG, balistocardiography (BCG), seismocardiography (SCG), photoplethysmography (PPG), respiratory rate measurement, skin conductance measurement (SC), etc. Normal values are described in the literature [12], and a deviation may indicate SI in a certain future time duration.

2. A Vision of the Future DMS Operation

The development of DMS began in the early 2000s, when Toyota implemented its system in the LS600h Lexus model. It uses a CCD camera with six infrared LEDs for active illumination of the driver to monitor facial cues. During this period, DMS relied on integrated sensors and software algorithms that monitor behavioral states such as fatigue, distraction, and drowsiness. These systems can identify head position, abnormal steering wheel, and pedal control behavior.

In the period 2010–2020s, modern DMS began integrating more advanced technologies, such as sensor fusion and AI-driven algorithms, mainly to support emerging Advanced Driver-Assistance System (ADAS) features in semi-autonomous vehicles. The systems during this period provided higher accuracy in assessing behavioral and psychological states based on head orientation and body posture, as well as the number of eye blinks and their duration, and eyelid closing percentage (PERCLOS) data.

In the last period, 2020–2024, contemporary DMS is an essential component of ADAS. They are now capable of robust eye-gaze vector estimation (X, Y, Z), detecting visual and cognitive distraction, attention state, impairment, fatigue, drowsiness, lane deviation, and alcohol intoxication. However, these systems still lack physiological monitoring for SI detection. For example, interior cameras can identify drowsiness based on closed eyes, but cannot differentiate between a sleepy driver and a driver who has lost consciousness.

Our vision of DMS operation with physiological monitoring capabilities is illustrated in Figure 1.

Additional biomedical sensors will be used to measure drivers’ physiological signals. If the measured signal’s parameters and features (shape, rate, time duration, etc.) are in normal ranges according to established standards, then the values of physiological parameters will be reported as normal. DMS will continue the monitoring without any indications or alerts for the driver. This normal and unobtrusive background monitoring is represented by the green cycle in the lower half of Figure 1.

However, when values deviate significantly or a long-term trend toward abnormality is observed, this may serve as a predictor of SI before medical event occurrence. In this case, the DMS generates a warning for the driver based on the detected predictor. Through integration with the Human Machine Interface (HMI) of the vehicle, the warning may be visual, audible, or both, prompting the driver to take appropriate action before the medical event occurs. This early warning gives the driver valuable time to respond adequately, avoid a potential road accident, and seek medical attention. This process is represented by the red/yellow colored cycle in the upper half of Figure 1.

The most valuable and informative physiological signals for reliable DMS operation are ECG, PPG, and RR signals. Capturing these signals allows for continuous determination of several vital signs. HR and HRV can be extracted from ECG and PPG. They are essential markers of the cardiac function and autonomous nervous system activity, which covers most of the potential SI events. SpO2 can be measured by obtaining PPG signals. From ECG and PPG signals, combined features can be extracted to estimate BP values. Using RR signal, it is possible to estimate the breathing rate and respiratory health of the driver. All three signals make the detection of early predictors available to the DMS for long-term real road driving.

3. Analysis of Sensors and Sensing Methods for Driver’s Physiological Parameters Monitoring

Research studies on DMS development have investigated 11 types of sensors classified into three categories according to the nature of the sensing element. A brief overview is provided in

Table 2. Sensors and their locations in the vehicle interior used in the research studies for driver physiological monitoring.

Sensor Category	Sensor Type	Physiological Parameter	Physiological Signal	Sensor Location Inside Vehicle Interior
Electro- magnetic	Contact electrodes	HR, HRV, EDA, BT	ECG, SC, BT	Driver’s body; steering wheel
	Capacitive electrodes	HR, HRV	cECG	Backrest; seat; armrest
	RADAR	HR, HRV, RR	HR, RR, BP	Backrest, windshield/rearview mirror; steering wheel; seat belt
	Magnetic induction	HR, RR	Bioimpedance	Backrest; seat belt
Mechanical	Accelerometer	HR, RR	BCG, RR	Backrest; seat belt
	Piezo-electric	RR	RR	Seat belt
	BCG sensor	HR	BCG	Backrest; seat
Optical	Visible light video camera	HR, HRV, RR	iPPG	Windshield/rearview mirror; central console; steering wheel
	Infrared light video camera	HR, HRV, RR	iPPG	Windshield/rearview mirror; central console; steering wheel
	Pulse oximeter	HR, HRV, SpO2, BP	PPG	Steering wheel
	Smart watch, wrist band, smart ring with built-in sensors	HR, HRV, SpO2, EDA	PPG, SC	Driver’s body

* Gold standard; Low-reliability technique; Low accuracy; High accuracy.

. A color code is used to clearly differentiate between them. Dark blue highlights the sensors that have been studied but have not met the expectations in terms of accuracy and reliability in the automotive environment. Light blue marks low-accuracy sensors, but scientific interest is still ongoing. Green indicates the sensors that offer high accuracy and reliability for long-term real road monitoring. These two sensors are often used as a benchmark (gold standard) for comparing accuracy with other sensor measurements, but are considered inconvenient for the driver and not suitable for everyday use. Orange highlights the most advanced and widely adopted sensors in recent years. They provide higher accuracy in real road monitoring scenarios and are unobtrusive for the driver. Recent scientific efforts aimed at developing more robust techniques are constantly improving their reliability in vehicle interiors.

Future work in the field of physiological monitoring in DMS will focus on developments using RADARs, visible and IR light cameras, and wearable devices. Other sensors may still be utilized, but primarily in a supporting role. For example, sensor fusion techniques can combine data from an accelerometer and RADAR. The accelerometer signal is capable of capturing motion-related artifacts, which enables the separation and suppression of noise components present in the RADAR signal data.

Over time, a number of physiological monitoring methods for drivers have been developed, each with unique advantages and limitations. However, few studies have examined how well these methods perform in real-world driving conditions, with the majority of them being tested mainly in controlled laboratory settings using driving simulators. It is important to note that results from real-road experiments frequently support, complement, and validate those from laboratories.

Table 3. Summary of advantages and disadvantages of the sensing methods used in research studies for driver monitoring.

System Type	Sensing Method or Device	Advantages	Disadvantages
Wearable	Portable hub for physiological monitoring with contact electrodes	High accuracy during real driving Continuous long-term monitoring	Discomfort during wear in result of obtrusive contact electrodes
	PPG/smart watch PPG/wrist band	High accuracy during rest; easy to wear and use; affordable and widely available	Does not record data if not worn; issue due to need of recharging; lack of reliability in connection; low accuracy during physical activity
	PPG/smart ring PPG/neckless PPG/clothing	High accuracy; easy to wear continuously	No research data from studies Discomfort during wear
Non-wearable	ECG/steering wheel cECG seat cECG/safety belt	Unobtrusive techniques for the driver; measurements available at any time; no need of any driver involvement	Requirement of two-handed grip at precise steering wheel location; interference from triboelectric effects
	BCG/Accelerometers		Significant impact of motion artifacts; low accuracy
	MI Gradiometers		Not suitable for real driving conditions due to low accuracy
	iPPG/camera	Analyze facial expressions based on the driver’s face reference points; plus advantages	Accuracy is not guaranteed due to vibrations and ambient light; requirement of direct line of sight; personal data privacy is not protected; sensitive to skin tone variations and driver distance; reduced accuracy due to facial make-up and glasses
	RADAR	No need of direct visibility; operates in foggy and dark environments; operates through obstacles (clothing); privacy of personal data; monitoring of multiple objects; plus advantages	Requires precise positioning to minimize motion artifacts; needs more accurate algorithms for determining physiological parameters; interference from nearby electronic equipment

summarizes the key advantages and disadvantages of the sensing methods.

Devices for physiological monitoring can be classified into two main categories: wearable and non-wearable.

Wearable devices generally offer high measurement accuracy due to their relatively stable sensing area compared to non-wearables, which are more susceptible to motion-related artifacts. Smart watches and wrist bands are currently promising devices for real-world DMS applications among wearable technologies. These devices are widely accessible and frequently adopted by drivers as lifestyle accessories.

Notably, there is a research gap on the use of smart rings for physiological driver monitoring, as highlighted in green in Table 3. Although smart rings are less commonly adopted than smart watches, their sensing area tends to be more stable, potentially yielding higher measurement accuracy. Furthermore, smart rings are specifically designed for health monitoring. There are no extra features such as GPS or phone call functions, resulting in lower power consumption. Hence, they have a longer battery life, which enables extended monitoring.

Despite the advantages listed in Table 3, methods with wearable devices face challenges such as maintaining reliable communication with the DMS. Different communication standards may be a barrier to real DMS applications. User habits, including the risk of forgetting to wear the devices or recharge the batteries, pose an issue for the crucial requirement of continuous monitoring.

In recent years, non-wearable methods have seen significant advancements, driven by strong scientific interest. Their primary advantage is unobtrusive monitoring, which has become a dominant focus in both research and commercial DMS products. These systems do not require any active participation from the driver and are often perceived as additional vehicle features. However, integration of non-wearable sensors into interiors can be challenging due to limited space and interference from the presence of other electronic equipment. Nevertheless, these sensors can be incorporated during the production of the vehicle or installed afterwards as an additional device. Based on this model, leading companies in the DMS market, such as Smart Eye, Seeing Machines, and Cipia, are actively developing product solutions.

The accuracy of non-wearable methods continues to improve with advancements in hardware capabilities and sophisticated machine learning algorithms, progressively approaching the precision levels of wearable devices, especially when using several sensor modalities or more than one sensor from the same type for higher accuracy. For example, two or more cameras capture video of the driver from different interior locations.

Among non-wearable devices, camera-based methods have several inherent limitations, including optical path obstacles, vibration-induced noise, ambient light interference, and privacy concerns related to the monitored subjects (see Table 3). RADAR-based methods have fewer disadvantages. Ongoing research aims to improve accuracy and reliability by optimizing sensor placement within the vehicle interior to minimize motion artifacts. At the same time, semiconductor manufacturers are producing advanced RADAR system components operating at higher frequencies, thus reducing interference with other electronics components.

4. Conclusions

SI is one of the causes of fatal road crashes. The majority of them are due to cardiovascular problems. There is a high risk for the driver experiencing SI if he/she is not able to stop the car. Monitoring based on vital signs can provide detection of a potential life-threatening event. If a predictor for SI is detected, then valuable time for reaction is gained, and a vehicle collision may be avoided. Emergency medical conditions have the potential to be identified through physiological measurements. For example, cardiac arrest may be identified with ECG signal. In addition, abnormal vital signs, including HR, BP, and RR, are associated with several related conditions. A higher degree of abnormality can indicate a higher risk at the time or in the near future.

A system of biomedical sensors will monitor the driver’s health and generate a warning in case of SI. Comparison and analysis of present sensing methods and available sensors show the potential of electronic system development for accurate early detection of life-threatening medical events.

An additional direction of future work will focus on algorithms and mathematical models for digital processing aimed at extracting and determining the parameters and characteristics of the obtained bio-signals.