Pediatric epileptologist, Dr. Dave Clarke, is willing to navigate the promises and perils of telehealth in the name of delivering better diagnosis and management of epilepsy in children.
Science fiction often foretells both exciting and sometimes dystopian advances in technology that someday may become part of our everyday reality. From cell phones to smartwatches and from virtual reality headsets to self-driving cars, things that once seemed like pure fantasy are fast becoming seamless parts of our lives.
Today’s portable medical devices offer continuous monitoring, real-time alerts, machine learning (ML) analysis, and cloud connectivity to a virtual or remote physician available for immediate assistance. With continued advances in so many areas of medicine offering instant feedback, including real-time telehealth services, there is one global disease in particular for which new innovations offer exciting breakthrough opportunities to change the lives of so many. According to the World Health Organization (WHO), more than 50 million people worldwide have epilepsy; nearly 80 percent of them live in low- and middle-income countries. Additionally, an estimated 70 percent of people with epilepsy could be seizure free if properly diagnosed and treated. Given the latest advances in embedded IoT devices, what can we do about this problem?
Dr. Clarke has seen firsthand the hurtful consequences of uncontrolled or insufficiently monitored epilepsy.
“There are conditions like infantile spasms that should be considered a medical emergency, but that often go missed for months," Dr. Clarke says. "They need to be treated immediately or they can be devastating. If you could better define what may be a seizure or another type of movement in a safe and effective way, it would be helpful to everyone in medical management.”
Dr. Clarke has inaugurated Project Epidet with the focus on developing a device that can provide an early warning system for caregivers of epilepsy patients. Noble as the ambition is, Dr. Clarke is deeply aware of the level of complexity when it comes to the immediate detection and prediction of an epilepsy-related seizure.
Part of the problem, according to Dr. Clarke, is that there are many different types of seizures, including focal seizures that begin in one area of the brain but can spread. There are also tonic-clonic seizures that involve loss of consciousness and generalized muscle contractions. Absence seizures cause a patient to blank out or stare into space for a few seconds. Myoclonic seizures cause brief, jerking spasms of a muscle or a muscle group. Tonic seizures are characterized by sudden stiffness or tension in the muscles of the arms, legs or trunk, and atonic seizures cause sudden loss of muscle strength. A device that monitors generalized tonic-clonic seizures, which involves overall generalized movement, may not monitor more subtle movements that characterize different types of seizure. There are a few FDA technologies that monitor movement and may be sufficiently sensitive, and research is ongoing to find other ways, but they aren’t approved quite yet.
Another challenge is that when moving a product from the realm of health wearables ‘for entertainment only’ into medical monitoring, the FDA requires evidence of efficacy and safety. The complexity for achieving this goes up by an order of magnitude, which is one of the reasons why so few companies are pursuing portable medical devices. But Dr. Clarke is undeterred because seizures can, and should, be controlled. For About 70 percent of people living with epilepsy, anti-seizure medication offers a path to becoming seizure free. For the other 30 percent of patients that can’t be helped with medication, seizure prediction is vital and creates the opportunity to improve their lives.
Despite the many types of seizures, there are some common physiological characteristics that are common across all of them. For example, at the onset of a seizure, a patient’s brain activity increases and other reactions can occur including sweating, body temperature fluctuations, and changes in heart rate. Seizures can become life-threatening due to cardiac or respiratory complications – sometimes tragically leading to sudden unexpected death in epilepsy (SUDEP). This risk is particularly high for those who have frequent, unrecognized seizures.
Dr. Clarke and his team are focused on building a user-friendly device that is wireless, comfortable to wear, and suitable for use by non-medical personnel. It can also detect many - if not all - types of seizures, often when the patient has no knowledge or recollection of the event and alert the caregiver immediately. Through continuous monitoring and collection of data, the Epidet would be able to log data and build a history of seizure events. This data could then be used in the prediction of future events to alert the patient to take precautionary measures.
“Seizures may affect systems of the body other than the brain,” says Dr. Clarke. “With Silicon Labs, we are looking at autonomic features – skin galvanometry (skin conductance), temperature, and accelerometry (the measurement of movement) that affect sleep and other parameters to better predict - using data - when seizures are starting. Other things that get me excited include how we can use heart rate and other parameters to define different seizure types. A device like this could radically change the way we look at patients and could improve access to epilepsy management. We’ll find answers more quickly, which is to everyone’s advantage.”
Advances in semiconductor technology are seeing devices get miniaturized to ultimately become fully disposable. We can now offer extremely small devices with tiny electronic packages that include an application microcontroller unit (MCU), wireless connectivity, analog peripherals, general purpose I/O, integration of a variety of power sources, and the capability to interface a variety of sensors depending on the data one would like to collect.
In the case of the Epidet, sensors monitoring ECG and electrodermal activity (EDA), temperature, movement (falling or rest conditions), and heart rate and variability (HRV) can easily be connected to a system on a chip (SoC) that has enough memory to run advanced algorithms, handle secure connections to a mobile phone or gateway, provide state-of-the-art security to protect the device and user from intrusive activity, and operate for a reasonable number of weeks before needing to be recharged or replaced.
These small, low-power devices are opening new design avenues for innovators like Dr. Clarke and his team.
“The Epidet clearly must be kid friendly, and kid proof, to be effective,” says Dr. Clarke. “This is actually the first thing I think about when building a device – how patient-friendly is it? Another very important consideration is cultural sensitivity. Careful thought is being put into modeling the device so it can be scaled globally and be effective as an epilepsy management tool in all cultures. Solving these problems for only one segment of our population would be a disservice since it is truly our goal to improve outcomes globally.”
Cultural diversity is another significant challenge for medical device makers (especially so when hoping to enter international markets). Social context, familial preferences, and expectations can influence the uptake of wireless medical devices, not to mention anthropometric characteristics and language barriers. Personalized care and privacy are especially emotive topics when caring for children.
Globally, 80 percent of people with epilepsy live in low- to middle-income countries, so a device that is simple to wear, cost effective, and improves access to care has the potential to dramatically reduce the global burden of epilepsy.
Like continuous glucose monitors (CGMs) and holter patches, the Epidet needs to be small and flexible enough to not restrict a patient’s everyday movements but still collect the data required. It also needs the necessary components and battery capacity to get the job done. In some cases, the patch could even be disposable, lasting for a week or weeks before recycling or repurposing, taking the environment and cost into consideration.
A flexible patch can be more easily fixed to the body using adhesive and give better signal acquisition during periods of physical activity and exercise. On the contrary, having a more rigid patch allows for a larger, replaceable battery, and would likely be worn on the arm or chest with an elastic band. It would also be more prone to displacement during activity and may require readjustment or calibration of measurements during movement.
For this reason, a smaller more flexible form factor is ideal and warrants the use of an SoC such as Silicon Labs’ EFR32BG22, selected for this project, that offers an application MCU, Bluetooth 5.2 connectivity, ultra-low power consumption, compact size, and a range of peripherals for sensor integration -- all in a single device.
The next challenge is implementing an accurate and reliable detection algorithm. Sensors will collect ECG and EDA readings, monitor temperature and device acceleration, and process all this information locally on the device.
The BG22 selected for this project includes an ARM Cortex-M33 core that operates up to 76.8 MHz, contains 352kB Flash (512k variety available) and 32kB RAM. The device has a separate Cortex-M0+ driving the Bluetooth radio subsystem, which leaves the M33 and enough memory available to handle the application and data processing algorithms. This allows the Epidet to obtain the frequency and resolution of data required to detect a variety of seizure types, processing all the data locally on the device and utilizing the Bluetooth radio to send conclusive results and provide the requisite updates to the caregiver and cloud for logging and further analysis.
This means that everyone involved will benefit from the device: the patient, their family and/or caregiver, and the healthcare provider. For the patient, it’s protective and for the family/caregiver it provides reassurance that they will be able to intervene when necessary. For the provider, it will determine true seizure frequency and allow for expedited decision making and care management.
Another major design decision trades off size, power consumption, battery selection, and whether the device will be disposable, rechargeable, or potentially have a replaceable battery, battery type, and in what form factor.
Design criteria is in part driven by the power consumption of the BG22, as it runs both the application and handles periodic transmissions over an optimized low-power Bluetooth connection, combined with the power consumption of various sensors, components on the signal chain, and external EEPROM (electrically erasable programmable read-only memory) required for storing the data samples. The data acquisition makes up a large part of the power consumption budget. Clever techniques and optimization are needed to avoid continuously monitoring all the sensors by default and thus reduce the average power consumption of the device. It’s also necessary to provide a convenient way to recharge the battery in the device.
Key to maximizing the battery life of the Epidet, the BG22 draws 27uA per megahertz while active and operates at 1.2uA in the lowest sleep state. This power-optimized device offers five energy modes: EM0 (fully operational), EM1 (sleep), EM2 (deep sleep), EM3 (stop), and EM4 (shutoff). Depending on the mode, various architectural components are enabled or disabled, allowing the system to optimize the overall power budget of the device.
In addition, utilizing sensor interrupts and features of Bluetooth that minimize TX/RX time saves power overall. RFSense, another feature unique to the EFR32 wireless MCU family of devices, “wakes up” an MCU from EM2 or even EM4 power modes, allowing further power savings for the device.
With the increase in the use of wearables to measure the vitals of a person, and in some instances administering drugs into the body, the need to have top-notch security has never been so prominent. The BG22 offers state-of-the-art hardware cryptographic acceleration to safely protect and store the secret keys required to maintain a private Bluetooth connection to the gateway or smartphone. It also offers Secure Boot with Root of Trust and Secure Loader, ensuring an authenticated chain of trusted firmware that begins from an immutable memory (ROM). This prevents malware injection, rollback, and ensures that only authentic firmware is loaded and executed on the device. Finally, the BG22 provides Secure Debug with Lock/Unlock, allowing only authenticated access to the debug port protected by public-key cryptography.
The Epidet makes full use of the security features provided by the BG22 to secure the communication links and protect private user information. Its firmware can be securely updated over Bluetooth to swiftly correct any potential flaws in the product and to add new features and enhancements.
It’s human nature to question the consequences should the device be hacked, used improperly, or report bad data that might cause a dangerous condition to go unnoticed, especially when trialing a new technology on your loved one. Today’s technology seeks to address these concerns in the design stages. Beyond the technology application itself, what is sometimes more difficult to address is the human interaction with the data thereafter.
“Sharing information via wearables with the patient, and/or a parent or caregiver will be relatively easy,” says Dr Clarke. “But how far and who is involved in the data captured could be much more complex. Privacy concerns are paramount within all healthcare data we use to advance care.There are larger issues within society at large. When do you tell the insurer? Or if you are an adult with epilepsy, does the company that you work with have to know? The patient needs to remain at the center, and we will need to think through all aspects related to privacy and ethics.”
As we deploy more smart devices to collect and share data across the internet, we can begin to collapse distance to facilitate remote operations. Portable medical devices can liberate users from the constraints of cost, accessibility, and privilege on the proviso that you are comfortable sharing your data on the radio waves.
As an industry, we also need to consider how we can ensure and guarantee internet connection once a person is outside of the medical center’s care. Questions we need to answer are related to whether or not we rely on the patient’s phone, their Wi-Fi network, or whether we need to set up separate channels for remote management like developing a cellular gateway specifically for connected devices, for example, so that Bluetooth and Wi-Fi devices can have a direct connection over cellular.
Epidet seeks to address a common, yet highly complex neurological disorder to greatly improve the lives of children and adults around the world. It starts with seizure detection and bodily monitoring, but there’s nothing to stop such a project from expanding to collect environmental data at the patient’s residence to better understand potential triggers.