Well, good morning, everybody. Thank you, Dr. Thomas, for giving me such a wonderful opportunity to share our technology with this amazing group of audience. As Dr. Thomas mentioned, my research focuses on wearable ultrasound technology. What does that mean? I'm sure I don't have to preach to the choir, you know, what is ultrasound and how ultrasound is important for critical care, but what we are doing is to engineer next generation ultrasound technology. We want to make it wearable for continuous monitoring, hence free monitoring. I have a few examples to illustrate this technology. So to disclose, I have a startup called Subsonics trying to commercialize this technology. And I'm sure you are very familiar with wearables. Some of you may wear those Apple Watch, Fitbit, Garmin, Samsung wristband, so on and so forth. They can measure lots of things. Basically, those signals they can measure can be put into two categories, physical signals, chemical signals. So all the signals goes on and on, but I'm not going to go through the entire list because they are not my focus. Here, my focus is that existing wearables can only sense signals on the skin surface or, you know, shallow tissues. However, there are lots of more signals below the skin surface. In your central vasculatures, musculoskeletal, collagen and bones, tissue fluids, and most importantly, your visceral organs. And here's my hypothesis. Those signals in your deep tissues, central organs, have a stronger and fast correlation to the disease status and the progression than those surface signals. There's plenty of room underneath the skin based on this wearable platform. Then question boils down to, how can we access those deep tissue signals, central organs, from a wearable platform, continuously and non-emissively, right? So not too long ago, my group published a comprehensive review article trying to summarize all existing approaches that may allow you to access those deep tissue signals. There are six of those technologies, including biopotential mapping, vibration sensing, reverse endophoresis, electromagnetic antenna, impedance tomography, and ultrasonography. We compared those six technologies in three critical metrics, penetration depths, spatial resolution, and temporal resolution. Eventually, we find out that our favorite is ultrasound, because ultrasound can provide a very wide range of penetration depths, a very high spatial resolution, and a fast enough temporal resolution, so it allows us to capture those dynamic processes in the human body. That means by integrating ultrasound on this wearable platform, we can continuously and non-emissively monitor those deep tissue signals. Ultrasound adds a new sensing dimension to the existing wearable community. What you're looking at here is a schematic design showing the wearable ultrasound patch we published about now six years ago. So this device has around 100 transducers, 10 by 10 array. Each one of them can be individually controlled to transmit and receive ultrasound. And this is the exploded structure showing one transducer element, so it has those multi-layers to maximize those ultrasound transmission into the human body so that you got the highest signal-to-noise ratio. At the bottom here, those images showing how soft this patch is. It can be flexible, stretchable, and can be twisted. What do we do with this ultrasound patch? We first thought about imaging, imaging those infrastructure component. This is a whole field called non-destructive examination so that we can image those defects in your engine components, in your airplanes, in your cars, buildings, and bridges, right? So this is the image showing a ultrasound patch imaging on a non-planar surface of two defects that are non-parallel to each other. So this is a schematic. This is the imaging result in three dimension. Imaging results in two dimension. Everything match the schematic design perfectly. The most exciting thing about this technology is we can apply them to the human body to solve biomedical problems. So what are the biomedical problems we had in mind? Hypertension. Why hypertension? I'm sure you know blood pressure is the number one risk factor for cardiovascular disease that kill more people than all cancers combined. So this is the device on the neck operational scenario, and there are two vessels, carotid artery and the jugular vein. So those vessels are typically 1.5 to two centimeter below the skin surface. So this is the operational scenario showing schematically these oceanic beings transmitted into the human body, and it will have two reflections from those two vessel walls, anterior wall and the posterior wall. And as you can imagine, when does the pressure pulse comes in, those two pulses will actually shift back and forth, right? And by monitoring the distance between those two pulses, we can monitor the vessel diameter. Through a serious well-established equation and calibration, we can convert vessel diameter to blood pressure. And because we can transmit 5,000 pulses every second, that means we can measure blood pressure 5,000 times every second. So this is the blood pressure waveform we can get. You can see this is the continuous bit by bit, and on those blood pressure waveform, we can notice lots of interesting features. So this is a validation between this ultrasonic patch result and the commercial tenomial result. You can see this waveform is similar in morphology, and after calibration, those two numbers also are very close to each other. Ultrasound is a very powerful platform technology. It can sense many things. And one powerful thing it can do is a blood flow by Doppler. And this is a phase array on the neck, and we can use the phase array to focus and steer the ultrasonic beam so that you can do those Doppler image, you know, the blood flow, right? So this is, again, the schematic showing the blood flow measurement in the carotid and tracheal vein. So this is a color Doppler image showing the flow in the carotid artery. And after we know all those critical incident angles and the blood vessel angle, we can measure those blood flow velocity without any calibration. So what you measure is what you get. So from this spectrum Doppler, you can identify end-diastolic velocity, pixistolic velocity from this wearable ultrasonic patch from a commercial ultrasonic equipment, and we have done blend-out analysis to show the high correlation of those two measurement results. Most recently, we demonstrated this wearable B-mode imaging work. I think that's what attracted Thomas' attention. So I'm sure you know B-mode imaging is used in many divisions in the hospital. And for cardiologists, so those are the four standard views they use to diagnose your heart, including apical four-chamber view, apical two-chamber view, peristal long-axis, peristal short-axis views. So this is the schematics showing this cardiac anatomy. This is the wearable imaging result from the patch. This is the imaging result from the clinical ultrasonic probe. You can see the close, similar quality between them. So what I want to highlight here is this so-called peristal short-axis views. So this is the view cross-section in this direction. As you can imagine, this cross-section can move along this way, and you can get this so-called basal view, mid-cavity view, and apical cavity view. And if you superimpose all those views together, you got the 17-segment model, with each segment corresponds to one particular area in your myocardium. Why is this interesting? Because by monitoring each segment on your 70-segment model, you know what's happening with each piece of heart in your myocardium. Then you know which part of this myocardium has so-called regional wall motion abnormality, right? So if you know those regional wall motion abnormality, you know which part of those coronary vessel has blockage or has reduced blood flow. There have been clinical evidence to show that regional wall motion abnormality is about half an hour earlier than EKG to predict a heart attack. So this is a live video showing how this wearable ultrasonic patch works. I should have brought a device here, but if you're interested, you are welcome to visit my lab. So this is the patch on the chest of my student. We have this cross-array, two cross-array. They are two linear, and those two linear arrays are independent or subtle to each other so that we can provide those independent view of your heart simultaneously in real time. So those are the two views from those two linear arrays, and you can see those cardiac beating motion anatomical structure in real time. And what I want you to notice is that when the student has this patch on his chest, and when he moves around, you can see this imaging quality is not compromised. It's not like, does not have any motion artifacts, right? That's because this patch has very intimate conformal contact with the skin surface, so those motion artifacts are immune to our system. So what does this do? If you can do those imaging of your heart during motion, we immediately thought about those stress echocardiogram. I'm sure you know that in current clinical practice, stress echocardiogram can only imaging your heart before and after exercise. However, what is really critical is imaging your heart during exercise because some of those cardiovascular risks are hidden until you put the maximum stress, maximum exertion onto the cardiovascular system, right? So using our technology, we'll be able to imaging your heart during this entire process before, during, and after exercise. So this is a B-mode imaging of your heart throughout this entire process, and these are the M-mode images extracted from those B-mode images. You can see this is a long trace. You can identify lots of features from those M-mode images, and some of those structures, features, you think they are noises, but in reality, if you zoom in, they are all signals corresponding to your valve, closing, opening, and the breathing patterns, so on and so forth. So there's lots of signals you can try to make sense out of it. So wearable imaging is a platform technology. It gives you lots of information. If you show them to the clinicians like you, they will be overwhelmed. They will say, hey, what am I supposed to do with those images? So we came up with a machine learning model based on FCN32, trying to automate processing those images so that we can give actionable information to the busy clinicians, and in this particular case, we use FCN32 to segment left ventricle volume from this apical four-chamber view, and once we know those left ventricle volume, we can derive those three critical cardiac indexes, stroke volume, cardiac output, and ejection fraction, and this is the left ventricle volume waveform. It's very dense, lots of data, but if you zoom in, those are three critical moments. You notice that this left ventricle volume waveform is very, very different in morphology. Right after exercise, you see it's pretty triangular. After a few minutes, after exercise, you can see it has more features, including this diagonal notch, because when the heartbeat slows down, the entire cardiac structure has more time to complete this entire process. So from this left ventricle volume waveform, we can derive ejection fractions, stroke volume, cardiac output, and heart rate. So I don't have to preach to the choir how important those waveforms are, but I just want to emphasize a few things. So to the best of my knowledge, this is the first non-invasive continuous monitoring of those critical cardiac indexes waveforms. In clinics, people have to use this Swann-Gan-Castor to measure those values, and for Swann-Gan-Castor, based on the thermal dilution, it can only measure one number per heartbeat, one value per heartbeat. Here, we can do three, zero, 30 times measurements, 30 values per heartbeat, and this is non-invasive. So this is bringing lots of value to this critical care community. Ultrasound is a mechanical wave, so it allows us to measure the mechanical properties of the human tissue. And I'm sure you know, mechanical properties of this human tissue is related to a lot of diseases, such as cancer, cardiovascular disease, aging, fibrosis, so on and so forth. This is the entire field called biomechanics, pioneered by UC San Diego professor, Y.C. Fung. So in this case, we can use this wearable patch to three-dimensionally map those tissue mechanics. So this is the schematic showing the mechanism. You have the ultrasonic wave, and each tissue interface will have a reflection, and you compress the entire stack. And I'm sure you can imagine, some of the softer tissue will be compressed more, some of the richer tissue will be compressed less. And that means those peaks will be shifted, you know, back and forth with different distance. By correlating those peak position here with the peak position here, then we can know which layer of the tissue is softer, which layer of the tissue is more rigid. So in this way, we can, you know, map out this tissue stiffness in three-dimension because of this two-dimensional array. So this is a nice demo showing by the student by flying a stick in the lab. You can imagine when he was flying from the bottom, those, you know, bottom, those protein gets dehydrated, right, and it becomes stiffer, and time goes by, the stiff part at the bottom actually grows. You can see there's the interface nicely move, you know, upwards, you know, as he keeps flying. So how do you know your measurements are accurate? So this is the measurement from the World Ocean Apache in three-dimension compared with the clinical magnetic resonant elastography in three-dimension. As you can see here, those two mapping results are very, very close to each other in terms of this anatomical distribution and also the absolute numbers, right? You can see they have a little bit discrepancy because of those different, you know, magnets and different frequency of this radiation and also boundary conditions. So we apply this technology to the human body where those parts were prone to those musculoskeletal injury, such as your lateral shoulder, anterior forearm, anterior thigh, and the posterior calf. So this is the B-mode imaging along the depths. This is the stiffness measurement along the depths. You can see those, you know, stiffness measurements correlates with this B-mode imaging very well. Why do we need, you know, frequent continuous monitoring of the tissue stiffness? So we can start about the one application called delayed onset muscle soreness. So muscle soreness is basically based on those, you know, muscle fiber rupture after intense exercise and training, right? So this is not a big problem for you, but for those soldiers in the battlefield, for those elite athletes in the game, right? If they have over-training, then their performance will be compromised. We need to come up with a mechanism to provide timely feedback so that their training performance can be, the training outcome can be, you know, maximized. So this is the control experiment of the students doing eccentric exercise, followed by natural recovery, massage, and the thermal treatment. You can see the top part is the objective measurement of your muscle stiffness. And as you can see here, this is a period of doing the exercise. A few minutes after the exercise, the muscle stiffness already begun to increase. Three hours, six hours, one day, two days, this muscle stiffness become the highest. The bottom here is the subjective feeling of this muscle soreness. You can see here, the subjective feeling does not become the highest until three days after, right? So this is obviously delayed, and with our technology, we can provide a timely feedback to guide the training, maximize the outcome. Ultrasound is a mechanical wave, so it can measure blood pressure, blood flow, form images, measure tissue softness. How about the chemicals? There are lots of chemical molecules, reactions, information in your deep tissue, in your body. How can we access those chemical information on a wearable platform continuously and non-invasively? So this is a technique called photoacoustics. If you do not know the mechanism, it's basically a technique coupling light absorption and ultrasound image. So this is a schematic showing that we use a ray of laser. Laser, because laser has a strong light intensity, so it can penetrate way deeper into the human tissue. We use laser to shine those deep tissues, and as you can imagine, those molecules will absorb those light, those specific wavelengths of the light, and when those molecules absorb the light, it will heat up, and when it heat up, it will swell. When it will swell, it will emit ultrasound, and those ultrasound will propagate to the skin surface and eventually picked up by those transducers to form images. So this is the mechanism. The laser we use is called a WEXO, vertical cavity surface emitting laser, and here are those images of the device. So you can see that those yellow squares are those laser diodes. It can penetrate about two centimeters below the skin surface. So this is the patch, flexible, stretchable, and it can be twisted. So this is the patch during operation, and you can see that when you mechanically deform it, the performance of the patch is not compromised. We use a laser diode of 850 nanometer. Why this wavelength? Because this wavelength allows us to differentiate blood from those four common liquids that you can normally find in a human breast cyst. So this is the B-mode imaging of the phantom showing those different liquids. You can see B-mode imaging cannot really tell those chemical information, but with photoacoustics, you can see that you can clearly tell blood from those four common liquids. So probably some of you know that if you find those liquid in a human breast cyst is healthy, but if you find blood, then it's a strong indicator for breast cancer. So one in eight women will experience breast cancer sooner or later in their lifetime, and fortunately, this breast cancer is highly preventable and treatable if discovered early. So with our technology, we can empower the patients to do more frequent screening at home without a visit to the hospital. So this is the imaging of the human vein in the arm. So this is a 3D imaging result, and along different directions, you can see that we got those different cross-section at different slices. So this is the corresponding photoacoustic signal amplitude corresponding to those images. So it turns out that this signal amplitude is linearly proportional to the temperature. That means it allows us to monitor the core body temperature. Why core body temperature? Because the skin temperature, which is frequently measured in the hospital, is very susceptible to the room environment temperature, your sweating conditions, so on and so forth. Core body temperature is a much more accurate, much more stronger correlation to this internal hemodynamics information, cancer growth, so on and so forth. So this is the Fenton experiments to show us this result of core body temperature measurement. So you can see that the linear relationship is very, very high. Temperature, signal amplitude. So this is the mapping result. Warm, cold, somewhere in the middle is a gradient. You can see that we can do those two-dimensional, three-dimensional mapping of temperature distribution. And we inserted two thermocouples at those two locations. You can see the thermocouple reading and the photoacoustic reading are corresponding to each other very well. And thermocouples are invasive, and you can only measure two points, two locations. With our technology, we can do things non-invasively, and it can give you three-dimensional mapping. So this is a lot of information available. So what about ultrasonic pad is great? How are we going to transmit data to the back-end receiver, to the cloud? How are we going to power the device? So recently, we come up with a circuit, wireless circuit, that allows us to transmit those signal wirelessly to the cloud, to the doctor's office, so that the patient can wear this device and freely walk around. So this is the patch, this is the circuit, and the circuit has low power consumption. So some size lithium-ion battery can last about three hours. And this is totally rechargeable, and this lithium-ion battery lifetime can be expanded by lowering down this duty cycle, which means we can do less frequent measurement, say once every minute, or once every two minutes or five minutes. A critical challenge with this technology is that you can imagine, we put this patch on the human body, and when you move around, your target will actually shift, right? So it will be displacement between your sensor and your human tissue target. So this is an experiment done by the student when he put the patch on the neck, and when he rotated his head from minus 80 to positive 80, you can see that this carotid artery is actually displaced, shifting with the displacement of about one to two centimeter, right? So this is a huge displacement. And only when this carotid is directly underneath the sensor can you get this blood pressure waveform. And when this carotid is moved away, you only get those tissue interface noises, right? So this does not mean anything. So this is a significant challenge for those wearable ultrasound technology for deep tissue monitoring. So recently, we come up with this machine learning algorithm based on VGG-13 to automate the selection of those transducers so that this device will be able to track a moving target and output this continuous blood pressure waveform even during motion, right? So you can see that this algorithm, we automatically selected the best channel and give you a continuous waveform. And a good question you may ask is, ultrasound is a very powerful technology, it's very mature, used in the hospital at many places. What's your value proposition? What type of value can bring to this community? So I would say, ultrasound, especially this POCRS, point of care ultrasound technology is a very, very powerful function already. It solved a lot of problems, but it still has three critical limitations. First, this POCRS ultrasound, it's still manual. You still have to use your hand to hold it. And that requires an experienced user to use it very well. If this user is new, does not have much training, you may get the wrong results, right? This is so-called operator dependency. That's why someday, the first day of this meeting, there are lots of training tutorial courses of this ultrasound technology, right? Because everybody has different habits. They use scanning different location, different pressure. So I'm sure I don't have to preach to the choir. This is a grand challenge for this ultrasound technology. And because this requires experienced user, so this technology is very expensive. It's not a problem for the US. It's not a problem for some of these modern developed countries. But for low, middle-income countries, one ultrasonographer is taking care of 200 patients. And the second limitation with this technology is it's still one-time testing. Because this requires experienced user, this measurements diagnosis is only available in the hospital, and it only gives you a snapshot of what's going on in the body. If you do not do this measurement, if you walk outside the hospital, the doctors have no idea what's going on. And most importantly, this technique has to require the patient to be stationary. They have to remain still, right? And the third, this technology still require gel. Gel is water-based, is cold, is evaporating. You still have to reapply it, and it creates a problem for disinfection if you want to use this device across different patients. With our technology, this patch, you just have to slap it on the part of the body you want to monitor. It's hands-free. And those electronic scanning will be able to steer the ultrasonic beam electronically. The results will be interpreted by the AI algorithm. So it takes those experienced users out of the picture. So it lowers down the barrier for people to adopt this technology, and empowers the general, you know, civilian, general patient to monitor by themselves. So this is like operator-independent, and because of that, it's low-cost. Everybody can use it. And because the patch is portable, it can provide continuous monitoring at anywhere, anytime. It gives you a comprehensive picture of what's going on in your body. Do not lose any information, especially for those critical, transient moments of your disease development. And most importantly, this technology allows you to, you know, image to monitor during motion. As I, you know, explained in this stress echocardiogram, some of those, you know, diseases will remain hidden until you put the maximum stress exertion onto the, you know, your hemodynamic system. And this patch allows intimate, conformal integration with the skin surface, so it's gel-free. It does not require those, like, you know, disinfection or frequent reapply of the gel. So with that, I would like to conclude by acknowledging my students and, you know, postdocs. I'm learning from them every day. And also my collaborators who provided the critical support. And also, there's a funding agency. The size of the logo is roughly proportional to the amount of funding I accumulated over the years. Thank you very much.

wearable ultrasound

health monitoring

deep tissue

real-time imaging

AI integration

at-home diagnostics