Device Technology Advances Medical Practices
Medical device technology is advancing, with breakthroughs coming via AI, Radar, miniaturization, sensors, robots, and lower power consumption. We take you on a fantastic voyage through the technology that will produce the next revolution in medical practices and improve patient outcomes.
Improving patient outcomes, enhancing diagnostics or regulatory compliance are all critical goals for embedded design efforts in the medical device industry, and the advances in hardware and software development hold the promise of making these tasks easier, longer in duration, more secure, and with higher accuracy and reliability than the last generation of equipment. One company describes a wearable medical band designed to detect “conspicuous” irregularities in a patient’s continuous vital sign stream and advise an action to help, such as taking medication. However, it can also differentiate and detect “serious” irregularities, alerting doctors or emergency responders when triggered.
Behind the scenes at an embedded device manufacturer, the team should have a clear vision: Provide precision measurement of critical biometric parameters through a design with breakthrough capability and then attain regulatory approval before the competition. The resulting design would then allow medical professionals to gather patient data on the symptoms, conditions, and reactions to treatment. This is all to advance the process of diagnosing conditions and proposing appropriate treatments that address specific patient needs. In essence, the larger the data set that can be acquired from an individual through the collection of real-time biometrics and vital signs, the more accurate a medical staff can be in prescribing the most appropriate treatment, thus improving patient outcomes. Medical device design work seems straightforward and noble. Still, it presents several variables and dynamics that designers in other markets, like consumer electronics or industrial IoT, may not need to consider.
Regulatory compliance is one challenge that looms large here, stretching time-to-revenue expectations well beyond the comfort zone in most other markets and presumably raising stress levels for everyone involved. Why spend days, months, and years developing a system that may or may not pass approval? The medical device market is fast-growing, and some experts say it will grow 10 percent in annual revenues over the next ten years. Infineon states an estimated 64 percent growth rate in semiconductor sales for medical electronics. Then, there are the opportunities to advance medical science, improve humanity’s health, and make medical practices more effective. Additionally, there is the prospect of expanding the physical limits of system design beyond what’s possible now.
A new design might be able to measure pulse oximetry more accurately than a competitive one, but can the project meet the latest accuracy standards set by regulators? Perhaps more importantly, can it avoid the inaccuracy that might lead to misdiagnosis?
A design team might create a wearable medical device to be used by patients in a comfortable home environment to gather data more continuously, for example, but that design may not account for incidents seldom seen in a clinical setting. For example, if a patient is wearing a monitor for a heart condition, would the design account for unintended mishaps? If the patient turns in his sleep, will it tear off leads and possibly skew the readings that a doctor is recording at a remote location?
It’s interesting to see how embedded developers and OEMs address the unique challenges that medical device design demands to deliver exacting precision in the data they acquire and send to medical professionals. Unlike an IoT sensor mode in a factory gathering data and sending it off to the cloud, the variables and dynamics involved for the medical device developers make the task much more challenging. Still, there are basic design blocks that these devices have in common, such as those shown in Figure 1.
Infineon’s view of a wearable or disposable medical device block design. (Photo Credit: Infineon)
What are the most prominent challenges developers face in this fast-growing market? And what are OEMs proposing to help us all meet these challenges? There are three main areas where medical device technology is surgery, diagnostics, and drug discovery.
AI SUPER SENSOR
Nvidia built a platform called Holoscan [1] on its IGX reference design to allow device manufacturers to kick the tires on real-time AI applications. It also helps designers meet regulatory standards. To deliver real-time insights, Holoscan offers a bunch of sensor-processing tools to streamline the development and deployment of AI and high-performance computing applications. At the same time, they want to cut the time and costs required to build AI solutions.
Released in June, Holoscan is based on the Nvidia IGX Orin 700 and now supports the NVIDIA RTX6000 Ada GPU as a new configuration option. This GPU delivers up to 1,705 trillion operations per second, marking a 7x increase compared to an onboard iGPU. Holoscan in Figure 1 also supports a high-throughput network adapter called ConnectX-7 up to 200 GbE bandwidth and a GPUDirect RDMA path to GPU processing. Embedded security comes from an integrated module consisting of controllers to monitor critical operations, remote software updates and system recovery, and hardware root-of-trust.
The medical-grade reference architecture shown in Figure 2 comes with long-term software support that hastens the pace of innovation. The aim is to open the door for sensor innovation by processing high-throughput data streams. The platform promises to transform medical devices into continuous sensing systems to research and treat diseases.
Medical staff uses real-time insights from a high-performance AI computing platform based on the Nvidia Holoscan reference design. (Photo Credit: Nvidia)
According to a press release on Medtronic’s website, Kimberly Power, vice president of healthcare at Nvidia, is driving this program, which aims to push into the next frontiers in surgery by employing real-time AI devices. For surgery, diagnostics, and drug discovery, Holoscan provides AI to accelerate visualization capabilities and advanced computing to quicken the pace of new system introduction.
Nvidia offers long-life components and 10 years of software support, including IEC62304 documentation for software and IEC60601 attestation reports from embedded computing partners.
Medtronic is a giant in medical devices and wants to use AI as an assistant to healthcare professionals so that they can catch issues earlier and promote better patient results. The Nvidia Holoscan computing platform is one of the central pillars of this effort. Specifically, when searching for colorectal cancer, the new GI tool will reduce the miss rates of polyp detection, helping doctors with the early, real-time screening of colorectal cancer during colonoscopies [2]. Last year, Medtronic, Cosmo Intelligent Medical Devices, and Nvidia announced a collaboration called GI Genius to develop the industry’s first AI-powered endoscopy module.
Nvidia states that some of the largest medical device makers and dozens of robotic surgery and medical imaging startups are developing around the Holoscan platform. Nvidia’s platform offers pre-trained models and a microservices framework, allowing application deployments in an edge data server.
WEARING NOTHING
Early detection of possible health conditions is just one of the many areas that medical device component suppliers are targeting in the next wave of gadgets. Another important parameter is making devices less pervasive in patients’ everyday tasks. Making the collection of health-rated data more continuous allows a more accurate portrait of conditions that change throughout a typical day. That means stuffing sensors, actuators, and connectivity modules inside tiny form factors, so patients don’t even know they are wearing them.
What if they don’t have to wear them at all? A patient monitoring solution called the Millimeter Wave Radar Biometric Detection System by a company called Fingal Link [3] monitors vital signs continuously without direct skin surface contact, as shown in Figure 3. Made specifically for aged populations in nursing homes, this small box collects blood pressure, heart rate, and sleep quality throughout the day over the air. The key is Infineon’s 60GHz radar IC, which acquires data and securely sends it to the cloud.
Millimeter Wave Radar Biometric Detection System by Fingal Link gathers vital signs from patients without direct skin contact. (Photo Credit: Infineon)
Sitting somewhere in a patient’s room, this Radar IC emits radio waves, receives reflections from an object, and gets information. Developers have a choice of radar devices from Infineon, including the 24 GHz microwave variety and a 60 GHz millimeter wave option, which are capable of high detection accuracy and a wide frequency swing. Infineon says this non-contact collection of biometrics emphasizes privacy because it could replace systems that rely on continuous image-capture cameras.
GLUE TOOTH
While most healthcare wearables are designed for skin contact, Silicon Labs and Lura Health have teamed up to take readings from the saliva in a patient’s mouth.
Based on the Silicon Labs xG27 SoC family, which comes in sizes ranging from 2mm-squared to 5mm-squared, Lura Health is developing a device so small it can be glued to a tooth. Be careful flossing when it’s deployed. Data derived from saliva, doctors and clinicians intend to use the tiny device to test for more than 1,000 health conditions. [4]
Designed for the smallest IoT devices, the xG27 family ranges in size from 2 mm squared, about the width of a #2 pencil lead, to 5 mm squared, less than the width of a standard #2 pencil. These offer IoT device designers energy efficiency, high performance, trusted security, and wireless connectivity. This makes the xG27 SoC family ideal for tiny, battery-optimized devices like connected medical devices and wearables (Figure 4).
Silicon Labs’ Bluetooth SoC and MCU are for small form factor devices, such as Lura Health’s in-mouth monitor. (Photo Credit: Lura Health)
Lura Health, a medical device manufacturer and participant in the Alpha program for BG27 has already selected the new SoC to form the basis of its new in-development smart wearable. Unlike other wearables on the wrist or other external skin contact points, the Lura Health monitor goes in a person’s mouth. Specifically, the device is so small that it is glued to a tooth. With the device, dentists and other clinicians can collect important data from saliva, which is used to test for over 1,000 health conditions.
The SiLabs EFR32BG27C140F768IM32 chip is marketed as a complete IoT SoC, boasting Bluetooth wireless connectivity for portable and wearable medical devices. Its features include high-performance RF, low current consumption, a coulomb counter, and security compliant with SESIP L3. IoT device makers can create the smartest, fastest, and most energy-efficient applications while securing end-user privacy. Memory of up to 768 kB of Flash, 64 kB of RAM, and 18 GPIO provides maximum resources for software, designs, protocols, and peripherals while leaving room for growth.
According to the website on SESIP L3, PSA Certified is the independent security evaluation scheme for Platform Security Architecture (PSA) based IoT systems. It establishes trust through a multi-level assurance program for chips containing a security component called a Root of Trust (PSA-RoT) that provides trusted functionality to the platform. Before digging further, attackers should think twice before using this device as a backdoor to your health profile.
LAB IN POCKET
One new wave of medical device development is to build a closer relationship between sensor expertise and circuit design. Analog Devices touts its latest effort to marry electrochemical biosensor developers with measurement and testing system designers to develop a potentiostat.[5] The thinking behind this is that electrochemical biosensor developers have an easy time recognizing the electrochemical signature of a virus or bacteria. However, building products or readers like a potentiostat could take years of research.
The EmStat Pico potentiostat module by Analog Devices, Inc. (ADI) and PalmSens eliminates the need for specialized knowledge when creating electrochemical biosensor technology. By partnering with PalmSens, ADI attempts to fill in the knowledge gaps so innovators can get products to market faster. With ADI’s signal processing expertise, PalmSens created a potentiostat that is faster, smaller and has lower power than its laboratory equivalent without sacrificing accuracy.
Improved biosensing sensors and readers are supposed to give healthcare providers a quick aid to make the correct diagnosis and treatment—not only for COVID-19 but also for many other diseases ranging from malaria and tuberculosis to bacterial infections. The potentiostat circuit senses changes in a cell’s resistance and adjusts the current accordingly through an auxiliary electrode. This device is an essential piece of electronic hardware for control and measuring in most electroanalytical experiments.
This project also underscores the need to detect infections right at the point of care and eliminate the need for specialized office equipment, patient visits, and the delays in diagnosing a condition when you must send a sample to an offsite lab and wait a week or two to get the results back.
A use case ADI describes is from QSM, shown in Figure 5, which makes point-of-care veterinary diagnostics for rapidly identifying bacterial infections. Yes, that’s correct. The idea presented in the literature says that first, vets will use the device on dogs and cats, and then, the technology will eventually transition to human use. QSM set out to create an electrochemical biosensor and reader that provides diagnostics right in the office, with results available by the following day or even before your dog leaves the clinic. The instrument is fast and portable due to the embedded EmStat Pico form factor of 30 mm × 6 mm.
An electrochemical biosensor and reader that provides diagnostics in the office, with results available by the following day or even before your dog leaves the clinic. (Photo Credit: Analog Devices)
Just swab Fido’s ear put the sample in the pocket lab, and the results come quickly.
The heart of this device is the ADuCM355, an on-chip system that controls and measures electrochemical sensors and biosensors. The Arm Cortex-M3 processor is an ultralow-power, mixed-signal microcontroller. It features current, voltage, and impedance measurement capability. The ADuCM355 features a 16-bit, 400 kSPS, multichannel successive approximation register (SAR) analog-to-digital converter (ADC) with input buffers, built-in antialias filter (AAF), and programmable gain amplifier (PGA).
I, PILLBOT
What if there’s a deeper issue with a patient that goes beyond skin deep? There’s the recent PillBot announcement by Endiatx, which is designed for endoscopies. A patient then swallows this vitamin-pill-sized device at home, and then a doctor remotely operates the vessel with a game controller. After announcing in July, the creator of this little robot said it is now going through FDA clinical trials, which it expects to complete in 2025. Commercial launch is predicted in 2026.
A company press release states the benefits of a cost-effective, non-invasive tool for diagnostics and treatment in gastrointestinal care. Four out of five upper endoscopies yield negative results and show that there are no issues in the patient’s stomach requiring medical treatment. PillBot promises to spare patients from the discomfort of sedation and hospital visits. [6]
The device has three pumpjet motors to help it through yesterday’s lasagna (Figure 6). It also has a data transceiver for real-time wireless communication, a camera for transmitting live video feeds, LEDs to illuminate the stomach interior, a Lithium battery for 30-minute voyages, and a multi-layer flex circuit to condense components into the small sub. Version two of this tiny miracle will come with AI.
Endiatx’s cutaway image of Pillbot shows some inner workings of the vitamin-sized device. (Photo Credit: Endiatx)
Check out a video demonstration of this diagnostic tool from a recent TED Talk. A link is available on Circuit Cellar’s article resources page.[7]
BRAIN-TO-TEXT
This next innovative medical device promises viability in the long term. It is called the Miniaturized Brain-Machine Interface (MiBMI), a groundbreaking solution for individuals with severe motor impairments recently unveiled by researchers at EPFL in an article in the IEEE Journal of Solid-State Circuits.
The literature explains that MiBMI is a high-performance, compact, and low-power device that enhances the efficiency and scalability of brain-machine interfaces, making it suitable for practical, fully implantable applications. As shown in Figure 7, the two-chip solution measures only 2.46 mm2.
A miniaturized brain-machine interface proposed by EPFL could help patients with neuro-motor disorders communicate through brain-to-text conversion. (Photo Credit: EPFL)
This remarkable research promises to be a boon for people with motor control issues, such as those with ALS or spinal cord injuries. Past designs for this purpose have been impractical due to their bulky size and intense power demand. EPFL researchers say this new MiBMI is minimally invasive and integrates recording and processing on two small chips.
In a release [8] on its website, EPFL describes brain-to-text this way:
Brain-to-text conversion involves decoding neural signals generated when someone imagines writing letters or words. In this process, electrodes implanted in the brain record neural activity associated with the motor actions of handwriting. The MiBMI chipset then processes these signals in real time, translating the brain’s intended hand movements into corresponding digital text. This technology allows individuals, especially those with locked-in syndrome and other severe motor impairments, to communicate by simply thinking about writing, with the interface converting their thoughts into readable text on a screen.
In the IEEE journal abstract [9], the title describes this as “A 2.46-mm2 Miniaturized Brain–Machine Interface (MiBMI) Enabling 31-Class Brain-to-Text Decoding,” potentially extending speech synthesis and handwriting assistance to patients. EPFL says current BMIs rely on cumbersome benchtop setups with resource-intensive computing units, restricting their suitability for daily use.
EPFL describes the devices as capable of accurate, multiclass neural decoding and high-density sensing in a millimeter-scale silicon footprint. A 512-channel, 31-class neural decoder employs a novel concept of distinctive neural code (DNC) driven by a class saliency model. This facilitates the precise translation of intricate neural activity into handwritten characters using a low-complexity linear discriminant analysis (LDA) classifier.
The proposed decoder significantly improves memory utilization and computational complexity compared to a conventional LDA without DNCs. Moreover, MiBMI enables area-efficient 192-channel neural recording through time-division multiplexing, demonstrating its potential for fully integrated BMIs. Fabricated in a 65-nm CMOS process, the high-channel-count BMI chipset consumes 883μW. The proposed decoder translated human intracortical neural activity into 31 characters with 91.3 percent accuracy, significantly enhancing the task complexity compared to previous on-chip BMIs. Furthermore, MiBMI achieved 87 percent accuracy in decoding the neural responses of a rat to six classes of acoustic stimuli in an in vivo experiment.
The rat wanted some cheese in a text he sent to his spouse. No, I made the last part up, but this is a remarkable achievement that shows tremendous promise for the medical device arena and for patients with neuromotor conditions. Located in Switzerland, EPFL is one of Europe’s most vibrant and cosmopolitan science and technology institutions. According to Wikipedia, the letters stand for École Polytechnique Fédérale de Lausanne.
An EPFL states that current BMIs record the data from electrodes implanted in the brain and then send these signals to a separate computer to do the decoding. The MiBMI chips records the data but also processes the information in real time—integrating a 192-channel neural recording system with a 512-channel neural decoder. This neurotechnological breakthrough is a feat of extreme miniaturization that combines expertise in integrated circuits, neural engineering, and artificial intelligence. This innovation is particularly exciting in the emerging era of neurotech startups in the BMI domain, where integration and miniaturization are key focuses. EPFL’s MiBMI offers promising insights and potential for the future of the field.
— ADVERTISMENT—
—Advertise Here—
To be able to process the massive amount of information picked up by the electrodes on the miniaturized BMI, the researchers had to take a completely different approach to data analysis. They discovered that the brain activity for each letter, when the patient imagines writing it by hand, contains very specific markers, which the researchers have named distinctive neural codes (DNCs). Instead of processing thousands of bytes of data for each letter, the microchip only needs to process the DNCs, which are around a hundred bytes. This makes the system fast, accurate, and with low-power consumption.
REFERENCES
[1] Nvidia Holoscan https://nvidianews.nvidia.com/news/real-time-healthcare-industrial-scientific-ai-applications-igx-holoscan
[2] Medtronic embraces AI https://news.medtronic.com/why-NVIDIA-is-bullish-on-artificial-intelligences-future-in-healthcare-newsroom
[3] Infineon Millimeter Wave Guide Vital Detection system https://www.infineon.com/cms/en/about-infineon/make-iot-work/smart-health-appliances/
[4]Silicon Labs small form-factor SoCs
https://news.silabs.com/2023-03-14-Silicon-Labs-Announces-New-Bluetooth-SoC-and-MCU-Ideal-for-Small-Form-Factor-Devices
[5]Analog Devices and PalmSens build a potentiostat: https://www.analog.com/en/signals/articles/palmsens-enables-electrochemical-biosensor-poc-diagnostics.html
[6] Endaitx Pillbot release: https://endiatx.com/Press/PressReleaseJune/
[7]Endaitx TED talk: www.ted.com/talks/alex_luebke_vivek_kumbhari_how_you_could_see_inside_your_body_with_a_micro_robot?subtitle=en
[8]EPFL describes brain-to-text conversion https://actu.epfl.ch/news/an-entire-brain-machine-interface-on-a-chip/
[9] EPFL proposes a tiny brain-machine interface https://ieeexplore.ieee.org/document/10643873
RESOURCES
Nvidia | www.nvidia.com
Medtronic | www.medtronic.com
Infineon | www.infineon.com
Silicon Labs | www.silabs.com
Endiatx | endiatx.com
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • OCTOBER 2024 #411 – Get a PDF of the issue
Sponsor this Article
Tom Murphy has been plying his trade as a technology journalist for more than 25 years, first as an editor/writer/reporter and then as a communications professional. He has earned industry recognition for both journalistic endeavors and technology campaigns. Before joining Circuit Cellar magazine as editor-in-chief (2024), Tom wrote tech briefs, newsletters, blogs, press releases, white papers, and technical articles for numerous clients, primarily in the semiconductor industry. His introduction into the industry came as an editor for “Electronic News” covering microprocessor companies, the fabless semiconductor phenomena, communication chips, and many others. When he’s not tapping away at the keyboard, Tom learns Dungeons and Dragons terminology from his 10-year-old daughter and walks his bulldog (slowly) around the block.
