In hospitals and doctor's offices, medical ultrasound equipment is undergoing a dramatic change. Advanced ultrasound machines, long wheeled around on carts with various wires and probes hanging from them, are about to be permanently pushed aside, replaced by handheld probes that can send images to mobile phones. Small enough to fit in the pocket of a white coat, these devices have the flexibility to scan and image any part of the body, from deep organs to superficial veins, with a full 3D view using just one probe. At the same time, combined with artificial intelligence, these devices can be operated by untrained professionals in a variety of settings, not just by trained sonographers. In 2018, Butterfly Network, a company in Burlington, Massachusetts, launched the first of these small, handheld ultrasound probes. In September 2023, Exo Imaging, a company in Santa Clara, California, launched a competing version.

What makes this possible is silicon ultrasound technology: 4,000 to 9,000 transducers (devices that convert electrical signals into sound waves and vice versa) are crammed onto a 2 cm x 3 cm silicon chip using microelectromechanical systems (MEMS). By integrating MEMS transducer technology and complex electronics on a single chip, these scanners are not only able to reproduce traditional imaging quality and 3D measurements, but also open up new applications that were previously impossible.

01 Learn about ultrasound technology

Understanding the basics of ultrasound technology helps to understand how the researchers achieved this feat. An ultrasound probe uses a transducer to convert electrical energy into sound waves, which travel into the body and bounce back from the body's soft tissues, where they are picked up by the probe. The transducer then converts the reflected sound waves into an electrical signal, and a computer converts the data into an image that can be seen on a screen. Traditional ultrasound probes contain an array of transducers made from piezoelectric crystals or ceramic sheets such as lead zirconate titanate (PZT). When struck by an electrical pulse, these crystals or ceramic sheets expand and contract, producing high-frequency ultrasound waves that bounce off the inside. To help with imaging, the ultrasound waves need to travel outside the transducer and into the soft tissues and fluids of the patient's body. This is no simple task. Catching the echoes of these sound waves is like standing on the edge of a swimming pool and trying to hear someone talking underwater. Therefore, the transducer array uses multiple layers of material that smoothly transition from the hard piezoelectric crystal in the center of the probe to the soft tissues of the body. The frequency of the energy delivered into the body is primarily determined by the thickness of the piezoelectric layer. Thinner piezoelectric layers transmit higher frequencies, allowing smaller, higher-resolution features to be seen in ultrasound images, but only in superficial areas of the body. Thicker piezoelectric materials transmit lower frequencies, which can reach further into the body, but at lower resolution. Therefore, imaging different parts of the body requires several types of ultrasound probes, with frequencies ranging from 1 to 10 megahertz. To image large organs deep in the body or a baby in the womb, doctors use 1 to 2 megahertz probes, which have a resolution of 2 to 3 millimeters and can reach up to 30 centimeters deep into the body. To image blood flow in the carotid artery, doctors typically use 8 to 10 megahertz probes.

02 Multiple probes required

The need for multiple probes and a lack of miniaturization meant that traditional medical ultrasound systems were housed in heavy box-like machines that had to be wheeled around on a cart. The introduction of MEMS technology changed that. Over the past 30 years, manufacturers have used MEMS to create microscopic, extremely sensitive precision components for many industries (see "Inkjets that do more than print paper"). This advance has enabled the creation of high-density transducer arrays that produce a full range of frequencies from 1 to 10 megahertz, allowing imaging of tissue at various depths within the body with just a single probe. MEMS technology has also helped shrink other components so that they can all fit on a handheld probe. Add in the computing power of a smartphone, and there's no need for a bulky cart. The first MEMS-based silicon ultrasound prototypes appeared in the mid-1990s, when excitement about MEMS as a new technology was at its peak. The key component of these early transducers was a vibrating micromachined membrane that produced sound waves by vibrating the device in the same way that a drum would be struck in the air. Two architectures emerged. One of these is called a capacitive micromachined ultrasonic transducer (CMUT), so named for its simple capacitor-like structure. Pierre Khuri-Yakub, an electrical engineer at Stanford University, and his colleagues demonstrated the first CMUT devices. CMUTs are based on the electrostatic forces in a capacitor, which consists of two conductive plates with a certain gap. One of the conductive plates (the micromechanical membrane mentioned above) is made of silicon or silicon nitride with a metal electrode; the other is usually a thicker and more rigid MEMS silicon wafer substrate. When a voltage is applied, opposite charges are placed on the membrane and substrate, and the attractive force pulls the membrane toward the substrate and bends it. When an oscillating voltage is applied, the electrostatic force changes, causing the membrane to vibrate, just like hitting a drumhead.

When the membrane comes into contact with the human body, the vibrations transmit ultrasonic frequency waves into the tissue. How many ultrasound waves can be generated or detected depends on the gap between the membrane and substrate, which cannot exceed 1 micron. Microelectromechanical systems technology makes this precision possible. Another architecture based on MEMS technology is called a "piezoelectric micromachined ultrasonic transducer" (PMUT), which works similarly to a miniature smoke alarm buzzer. These buzzers have two layers: a thin metal disk fixed to the periphery, and a thin, small piezoelectric disk glued to the top of the metal disk. When a voltage is applied to a piezoelectric material, its thickness repeatedly expands and contracts. Since the lateral dimensions are much larger, the diameter of the piezoelectric disc changes more significantly, bending the entire structure in the process. In smoke alarms, the diameter of these structures is usually 4 centimeters, and the frequency of the alarm sound they produce is about 3 kilohertz. When the diameter of the membrane is reduced to 100 microns and the thickness is 5 to 10 microns, the vibration frequency rises to MHz and can be used for medical ultrasound. In the early 1980s, Honeywell developed the first piezoelectric film MEMS sensor on a silicon diaphragm. It was not until 1996 that Paul Muralt, a materials scientist at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland, produced the first PMUT device at ultrasonic frequencies.

03 A major challenge facing CMUT

A major challenge facing CMUTs is how to generate enough pressure to transmit sound waves deep into the body and receive the returning echoes. The extremely small gap between the membrane and the substrate restricts the movement of the membrane, which in turn limits the amplitude of the sound waves generated. Combining an array of CMUT devices of different sizes into a single probe can also expand the frequency range, thereby reducing the pressure output because the available probe area for each set of frequencies is reduced. Curry-Jacob's laboratory at Stanford University solved these problems. In the early 2000s, researchers in the laboratory discovered in experiments that if the voltage on the CMUT-like structure was increased, the electrostatic force would exceed the restoring force of the membrane, causing the center of the membrane to collapse onto the substrate. At first, the membrane collapse seemed to be a complete failure, but it was later proved that this method could both improve the efficiency of the CMUT and help adjust the device to different frequencies. The efficiency is improved because the gap around the contact area is very small, which can enhance the electric field there; the pressure is increased because there is still a large range of movement in the larger annular area at the edge. In addition, the frequency of the device can be adjusted by simply changing the voltage. This enables a single CMUT ultrasound probe to efficiently generate the ultrasound frequency range required for medical diagnosis. Since then, it has taken more than a decade to understand, simulate, and fabricate the complex electromechanical behavior of CMUT arrays. The process of simulating these devices is tricky because there are thousands of individual thin films interacting with each other in each CMUT array. On the manufacturing side, challenges include finding the right materials and developing the processes needed to produce smooth surfaces and ensure consistent gap thickness. For example, the thin dielectric layer that separates the conductive film and the substrate must withstand about 100 volts at a thickness of 1 micron. If this dielectric layer has defects, charge can be injected and the device can short at the edges or short where the conductive film touches the substrate, causing damage to the device or degraded performance.

Eventually, though, MEMS foundries like Philips Engineering Solutions in Eindhoven, the Netherlands, and Taiwan Semiconductor Manufacturing Company (TSMC) in Hsinchu, Taiwan, developed solutions to these problems. Around 2010, these companies began producing reliable, high-performance CMUTs. Early PMUT designs also struggled to generate enough pressure for medical ultrasound. But they can work in some consumer applications, such as gesture-detection sensors and distance sensors. In these “air ultrasound” applications, bandwidth is not critical, and frequencies can be below 1 megahertz. The introduction of large 2D matrix arrays for fingerprint sensing in mobile phones in 2015 gave an unexpected boost to the use of PMUTs in medical applications. In the first demonstration of this approach, researchers at the University of California, Berkeley and the University of California, Davis, attached about 2,500 PMUT elements to metal-oxide semiconductor (CMOS) electronics and placed them underneath a layer similar to silicone rubber. When a fingertip presses against the surface, the prototype measures the amplitude of the reflected signal at a frequency of 20 megahertz, identifying the ridges and grooves between the ridges of a fingerprint. This demonstration of integrating PMUTs and electronics on a silicon chip is impressive, showing that large 2D PMUT arrays can generate frequencies high enough for imaging shallow features. However, to make the leap toward medical ultrasound, PMUT technology requires greater bandwidth, greater output pressure, and more efficient piezoelectric films.

Some semiconductor companies have helped, such as Geneva-based STMicroelectronics, which has found ways to integrate thin films of lead zirconate titanate on silicon membranes. Additional processing steps are required to maintain the performance of these lead zirconate titanate films. But the performance improvement is worth the cost and the additional steps. To achieve a larger pressure output, the piezoelectric layer needs to be thick enough so that the film can withstand the high voltage and produce good ultrasound images. But increasing the thickness makes the membrane stiff, which reduces the bandwidth. One solution is to use an elliptical PMUT membrane, which effectively combines several membranes of different sizes into one. This is similar to changing the length of a guitar string to produce different tones. Different parts of the elliptical membrane are wider or narrower, so a variety of string lengths can be provided on the same structure. In order to effectively vibrate the wide and narrow parts of the membrane at different frequencies, the solution applies electrical signals to multiple electrodes placed in corresponding areas of the membrane. This approach allows the PMUT to work effectively over a wider frequency range.

04 Histrory of CMUT

In the early 21st century, researchers began to push medical ultrasound CMUT technology out of the laboratory and achieve commercial development. Stanford University spun off several startups targeting this market. Leading medical ultrasound imaging companies such as General Electric, Philips, Samsung and Hitachi also began to develop CMUT technology and test CMUT probes.

But it wasn't until 2011 that the commercialization of the CMUT really began to gain traction. That year, a team with experience in semiconductor electronics founded Butterfly Networks. The IQ probe, launched in 2018, was a game-changer: the first handheld ultrasound probe that could image the entire body using a 2D imaging array and generate 3D image data. The probe, about the size of a TV remote control and only slightly heavier, was initially priced at $1,999, one-twentieth the price of a full-size cart-mounted ultrasound device. Around the same time, Hitachi in Tokyo, Japan, and Kona Medical Technology (Suzhou) Co., Ltd. in China commercialized CMUT probes for use in conventional ultrasound systems. But neither company’s products performed as well as Butterfly Networks’. For example, the CMUT and electronics were not integrated on the same silicon chip, which meant that the probe used a 1D array rather than a 2D array, a limitation that prevented the system from generating 3D images, which are necessary for advanced diagnostics, such as finding bladder capacity or seeing synchronized orthogonal views of the heart. In September 2023, Exo Imaging launched the handheld probe Exo Iris, marking the first commercial appearance of a PMUT medical ultrasound device. Developed by a team with extensive experience in semiconductor electronics and integration, the Exo Iris is about the same size and weight as Butterfly Networks' IQ probe. At $3,500, it's not much more expensive than Butterfly Networks' latest model, the IQ+, which costs $2,999.

The microelectromechanical system ultrasound chip in these probes is 2 cm by 3 cm in size and is one of the largest silicon chips with electromechanical and electronic functions. Its size and complexity bring production challenges in terms of product consistency and yield. These handheld devices have low operating power consumption, and the probe battery is relatively light. When connected to a mobile phone or tablet, the device can be used continuously for several hours and has a short charging time. In order to make the output data compatible with mobile phones and tablets, the main chip of the probe is digitized, and signal processing and programming are performed. To provide 3D information, these handheld probes scan multiple 2D slices of human tissue structures and then use machine learning and artificial intelligence to construct the necessary 3D data. Built-in artificial intelligence algorithms can also help medical staff place needles precisely in target locations, such as hard-to-find vasculature or other tissues that need biopsy. A 2022 study in NEJM Evidence, a subsidiary of the New England Journal of Medicine, showed that the artificial intelligence functions developed for these probes are so good that professionals without ultrasound training (such as midwives) can also use portable probes to determine the gestational age of the fetus with an accuracy similar to that of trained ultrasound physicians. AI-based capabilities could also enable the handheld probe to be used in emergency medicine, low-income settings, and for medical student training.

05 Small Ultrasound Equipment

Several of the world's largest semiconductor foundries, including TSMC and STMicroelectronics, are now producing MEMS ultrasound chips on 300mm and 200mm wafers, respectively. In fact, STMicroelectronics recently opened a dedicated thin-film piezoelectric MEMS production laboratory Lab-in-Fab in Singapore to accelerate the transition from proof-of-concept to mass production; Philips Engineering Solutions manufactures CMUTs for CMUT-on-CMOS integration; and Vermon in Tours, France, designs and manufactures commercial CMUTs. This means that startups and academic groups now have access to the foundational technology to achieve a higher level of innovation, and the cost is much lower than it was a decade ago. With these advances, industry analysts expect MEMS ultrasound chips to be integrated into many different medical devices for imaging and sensing. For example, Butterfly Networks is working with Forest Neurotech to develop MEMS ultrasound devices for brain-computer interfaces and neuromodulation. Other applications include low-power devices that can be worn for a long time, such as heart, lung and brain monitors, and muscle activity monitors for rehabilitation therapy. In the next five years, we can expect to see miniature passive medical implants with microelectromechanical system ultrasound chips that use ultrasound to remotely drive power and data transmission. Eventually, these handheld ultrasound probes or wearable arrays can be used not only to image human tissue structure, but also to read vital signs, such as changes in internal pressure caused by tumor growth or deep tissue oxygenation after surgery. Sensors similar to ultrasonic fingerprints can also be used to measure blood flow and heart rate. Perhaps one day, wearable devices or implantable devices will be able to generate passive ultrasound images in our daily life scenes such as sleeping and eating.

View more at EASELINK