Photoacoustic Imaging: From Organelles to Cancer Patients / Seminar Day, Session III

Summary

Highlights

• Photoacoustic imaging bridges light and sound for revealing intricate details of tissues. 🔦🔊
• It offers a non-invasive, multi-scale approach to view tissues from the microscopic to organ level. 🔬🧠
• The seminar dives into how this technology can revolutionize cancer detection and treatment. 🎯
• Dr. Wang discusses its potential for use in brain imaging, enhancing understanding of brain functions. 🧠
• The technique is advancing towards becoming a routine clinical tool due to its precision and safety. 🩺

Key Takeaways

• Photoacoustic imaging uses light to generate sound for sharp imaging of tissues. 🎇
• The technology offers both optical contrast and ultrasonic resolution for detailed insights. 👀
• Photoacoustic tomography can non-invasively image biological tissues, from organelles to whole organs. 🏥
• This method is particularly beneficial for cancer diagnosis, offering safer and faster imaging. 🎗️
• Future applications include non-invasive pathology and monitoring of physiological responses. 🔍

Overview

Chapters

• 00:00 - 00:30: Introduction and Opening Remarks The chapter begins with a nostalgic reference to the 1966 sci-fi movie 'Fantastic Voyage', celebrating its unique concept of using nanohumans for internal body repair. The mention of Raquel Welch adds a pop culture element, before transitioning to a discussion about advancements in 2021 concerning Dr. Li Hong Wang, a notable figure in the field of medical science. The chapter sets the stage for exploring innovative medical technologies by drawing parallels between past fictional ideas and current scientific developments.
• 00:30 - 01:30: Photoacoustic Imaging Overview The chapter introduces photoacoustic imaging as a non-invasive technique that allows deep visualization into biological tissues. It highlights the use of 3D photoacoustic microscopy and computed tomography as significant tools in diagnosis, disease monitoring, and surgery. The chapter seems to be a welcoming introduction to the technology ranging from organelles to its application in cancer patient management. Dr. Wang is acknowledged for his contribution to this field.
• 01:30 - 02:30: Motivation for Photoacoustic Imaging The chapter titled 'Motivation for Photoacoustic Imaging' discusses the multi-scale approach required for future biology and medicine. It emphasizes the importance of analyzing different length scales and integrating this information to derive solutions for complex problems. The chapter highlights how photoacoustic imaging is utilized in achieving this objective, extending its application from organelles to cancer patients.
• 02:30 - 05:00: Optical and Photoacoustic Imaging Techniques Optical imaging and photoacoustic imaging are pursued due to their unique ability to provide molecular specificity. The fundamental interaction between light and matter at the molecular level allows these imaging technologies to probe molecules, which is crucial given the fundamental roles molecules play in biology and medicine.
• 05:00 - 07:30: Photoacoustic Tomography and Its Challenges Photoacoustic tomography presents various challenges but remains crucial due to its capabilities in non-invasive imaging. It allows for in vivo functional imaging analogous to functional MRI, metabolic imaging similar to PET, molecular imaging of gene expressions, disease markers, and histological imaging without labeling.
• 07:30 - 10:30: Applications and Impact of Photoacoustic Imaging The chapter delves into the challenges posed by turbidity and opacity in various materials, noting that most substances in the world are not fully transparent. It emphasizes the reliance of humans on vision, a form of optical imaging, for understanding the universe. This sets the stage for discussing how these challenges impact the field of photoacoustic imaging.
• 10:30 - 13:30: Human Breast Imaging and FDA Approval The chapter discusses the challenges of optical imaging in biological tissues, which are generally turbid or opaque. To penetrate deeply into these tissues, one must overcome significant difficulties, such as the dispersion of photons within a short distance due to multiple scattering.
• 13:30 - 16:30: Brain Imaging and the Future of Photoacoustic Technology The chapter titled "Brain Imaging and the Future of Photoacoustic Technology" starts by discussing the concepts of scattering and diffusion, highlighting the similarity to heat diffusion. It mentions the mathematical modeling of these phenomena using the diffusion equation. The chapter also provides a historical perspective on optical imaging, tracing back to approximately 350 years ago with the invention of conventional microscopy, which enabled the examination of cells and subcellular structures.
• 16:30 - 19:30: Microscopic and Molecular Imaging The chapter discusses the progress in microscopic and molecular imaging techniques, starting from the optical aberration limit, which allowed imaging penetrating tens of microns. This limit was a barrier for hundreds of years until the invention of the laser. The invention of laser technologies, utilizing the coherence properties of modern light sources, helped overcome this barrier, reaching the optical diffusion limit. Unfortunately, this new limit allows penetration up to only about a millimeter in scattering biological tissues.
• 19:30 - 23:30: Real-Time and High-Speed Imaging The chapter discusses the advancements in imaging technology, particularly in overcoming the diffusion limit in tissue imaging. The invention of photoacoustics, or photoacoustic imaging/tomography, has been a significant breakthrough. This technology allows for penetration of imaging in biological tissues up to multiple centimeters with high resolution, a substantial improvement from previous limits which were near millimeters.
• 23:30 - 25:00: Future Directions and Conclusion The chapter discusses the concept of multiply scattered photons, explaining that while these photons are not absorbed or destroyed, their paths are altered. This scattering reroutes photons rather than eliminating them. By leveraging these scattered photons, it is still possible to achieve optical contrast. However, a method is needed to retrieve spatial information for high-resolution imaging. Therefore, the chapter explores the integration of light and sound to form a single imaging modality. It acknowledges a significant challenge known as the dissipation limit and indicates ongoing efforts to overcome this issue. The focus is on advancing the field of photoacoustic imaging.
• 25:00 - 31:30: Q&A Session with Dr. Wang The chapter provides an insight into the history and development of photoacoustics, a branch of science connecting sound and light. Dr. Wang describes how photoacoustics have been recognized for over a century, with Alexander Graham Bell being one of the pioneers in this field through his invention, the photophone. The photophone used sunlight to transmit sound, marking an early exploration of the relationship between light and sound before the advent of lasers.

Photoacoustic Imaging: From Organelles to Cancer Patients / Seminar Day, Session III Transcription

• 00:00 - 00:30 i'm going to go pop culture first on this session some of you may remember a 1966 sci-fi movie called fantastic voyage some of you may remember it for raquel welch in a latex bodysuit but what i remember it for is the radical idea of using nanohumans to fix the human body from the inside so now we come to 2021 and dr li hong wang the brain professor of medical
• 00:30 - 01:00 engineering and electrical engineering his work is allowing us to non-invasively peer deep into biological tissues and deliver 3d photo acoustic microscopy and computed tomography these are extraordinary tools for diagnosis for disease monitoring and for surgery so we welcome photo acoustic imaging from organelles to cancer patients and dr wang thank you pat well thank you all for joining me um
• 01:00 - 01:30 we'll be talking about photo acoustic imaging from organelles to cancer patients the future of biology of medicine is multi-scale we got to look at at different length scales and recombine all the information together to make sense to come up with fundamental solutions to all the problems we face and we work on photo acoustic imaging for this purpose
• 01:30 - 02:00 why do we work on optical imaging or photo acoustic imaging we have a number of imaging technologies already optical machine is very unique because we have molecular specificity from the physics perspective we know light matter interaction occurs at the molecular level this is a very unique position for us to probe molecules given the fundamental roles of molecules in biology and medicine
• 02:00 - 02:30 we really have to work on this problem despite all the challenges we face by detecting molecules we can already provide in vivo functional imaging very much analogous to functional mri in vivo metabolic imaging similar to positron emission tomography or pet in liver molecular imaging of gene expressions or disease markers even in vivo his stuff histological imaging without labeling
• 02:30 - 03:00 what challenges do we face if you look around the world almost all materials are turbid or opaque with the exception of maybe water and air even those materials are curbed to some extent if you look over long range then you'll find turbidity we understand the universe through observation humans observe through vision vision is actually a kind of optical imaging and so this is this tells us how
• 03:00 - 03:30 important optical imaging is if you zoom into biological tissues nearly all tissues are turbid or opaque so we have to conquer this challenging problem to gaze deep into biological tissues you can see here if you launch a photon pulse a very short pulse into a piece of biological tissue within about a millimeter the photons will get dispersed all over the place through a process called multiple
• 03:30 - 04:00 scattering or diffusion and you can see here the later stage of propagation is very similar to heat diffusion or diffusion and then we actually mathematically solve a similar equation called diffusion equation to mathematically model this problem if you look at the history of optical imaging you start from about 350 years ago when conventional microscopy was invented allowed us to gaze into a cell and look at subcellular structures but it doesn't
• 04:00 - 04:30 allow us to go very deep it penetrates about tens of microns it stops at the so-called optical aberration limit and we stagnate at this level for hundreds of years until laser was invented using the coherence coherence properties of modern light sources and we conquered this limit and then reached the next limit which is called optical diffusion limit and this is unfortunately only about a millimeter in scattering biological
• 04:30 - 05:00 tissue again we basically stopped at this level for decades until photoacoustics for the acoustic imaging or tomography was invented we overcame the diffusion limit advance the penetration by nearly two orders magnitude now we're talking about not even millimeter we're talking about multiple centimeters of penetration in biological tissue and still provide you very high resolution the key idea is that even though photons
• 05:00 - 05:30 are multiply scattered they're not absorbed they're not dead so scattering only re-route photons they don't kill photons so if we can leverage those photons we can still get optical contrast but we need to find a way to retrieve the spatial information for high resolution imaging and that's why we combine light and sound to form a single image modality we do face the next challenge which is called the dissipation limit and we're looking for ways to overcome the next challenge as well today i'll focus on photo
• 05:30 - 06:00 acoustic tomography photo acoustics as a physical phenomenon uh has been around for over a hundred years in fact alligator alexander graham bell first reported photo acoustics through his idea of photophone instead of telephone he encoded sound into a sunbeam there was no laser yet and the beam was propagated to the distal end and then the sound the light was
• 06:00 - 06:30 converted back into sound again through the photoacoustic effect now we have a set of modern concepts laser ultrasonics tomography computer none of those existed in bell's time and we want to combine them with this old physics to form a novel image modality called acoustic tomography in photo acoustic imaging we start with a very short pulsed laser
• 06:30 - 07:00 nanosecond laser pulses we expand the beam such that energy per area is within the safety limit we allow photons to wander around so we can provide deep penetration when photons absorb it generates very rapid heating every milli degree gives us a detectable signal and so photo acoustic sensing can be very sensitive we're talking about hundreds of pascals and the sensitivity is even order
• 07:00 - 07:30 manually lower than that so within the safety dilemma you can heat up to hundreds of million degrees then you get a very good significance ratio to work with to form a sharp image unlike standard optical imaging we don't detect return the light we detect a returned acoustic signal because acoustic scattering in tissue is orders managed lower than optical scattering we can form a very sharp image however the contrast is from light absorption
• 07:30 - 08:00 so we're combining optical contrast with ultrasonic resolution for deep tissue imaging biological tissue is to ultrasound as water is too light you know we can see a glass of water if you have a spoon we can spot a spoon really nicely to form a sharp image ultrasound can do the same job but if you were to use pure pure ultrasound you fire ultrasound pulsing you listen to the echo you get mechanical contrast you don't see the molecules
• 08:00 - 08:30 you don't see the color so we want to see the molecules that's why we want to combine the two forms of energy we have to solve math there's something called the inverse readon transform oh not talking about the math here in detail but just to give you the key idea behind the math let's say we have a thunderbolt when you see the lining you got time zero when you hear the thunderbolt you have a time delay t1 for absorber you know that this lanyard
• 08:30 - 09:00 took place on this circle shell with a radius of the speed of sound times t1 now if you have three such measurements you can triangulate and pinpoint this thunderbolt all right so for a volumetric biological tissue we're essentially doing the same thing except now we don't have a single sound source we have a unknown volumetric sound source we want to figure out each voxel or each point in order to form a 3d image and so
• 09:00 - 09:30 oftentimes we need hundreds even thousands of detectors in instead of just three but you get idea from this analogy in 2003 we reported the first functional photo acoustic image also the first in vivo image in animals with intact scalp and skull this is non-invasive functional imaging very similar to functional mri by wiggling one side of the whiskers the
• 09:30 - 10:00 contralateral side of the brain was activated we can see that activation using this non-invasive technique this paper is one of the most cited papers now in the field of photo acoustics it really kicked off the growth of our field as you can see in this chart we're plotting the number of tags or papers in our field over the year since 03 we started exponential growth since 2010 the conference on photo acoustics has
• 10:00 - 10:30 become the largest at the 20 000 attendee photonics west now it's the largest by far it's still growing very fast there are about 20 companies open companies working on this technology so this is commercially available in fact early this year one of the companies has received fda approval for human breast imaging and but they are they are also providing
• 10:30 - 11:00 equipment for some animal imaging for drug discovery for basic biological research as well you can see on the market activity of our technology for the 2003 work we use the single element transducer just to demonstrate the concept but it's kind of slow it takes about 20 minutes to get a image of a slice a virtual slice of a piece of tissue or a small animal now we want to have a faster imaging
• 11:00 - 11:30 version you can see here this is a single impulse with a single laser shot we get an image of a cross section a 2d version then you can repeat very quickly to get a full body 3d image i'll quickly walk through what we did you start with the laser let's say this is for trunk imaging you start with a laser you send it through this conical lens to generate a hollow cone beam then we use a full ring this is why we
• 11:30 - 12:00 call this panoramic you want to look at the object from all possible angles when possible then that allows you to get the best image we follow something called the nyquist sampling you want to sample densely enough uh this is very much analogous to the temporal nyquist you know sometimes as you watch a movie you can see that the wheel turns backward because we're not sampling fast enough so spatially we want to avoid that as well we want to sample things enough that allows us to
• 12:00 - 12:30 get high quality images pre-amplification as we learn the engineering you want to amplify your signal before the noise creeps in so you maintain the signature noise ratio we use a one-to-one data acquisition so this allows us to get parallel detection hundreds of channels will give you signals simultaneously with a single shot you get a image then of course we use computer to reconstruct to form an image and display the image here's a close-up
• 12:30 - 13:00 of the key component you have this hollow cone beam the object is here and you have focusing that we call this ultrasonic focusing to give you a slice resolution and then you have the inverse readon transform to give you lateral or xy resolution the system can be easily reconfigured for brain imaging so here we use a diffuser to generate a solid beam instead of a hollow beam because for brain imaging this is the best way to use your laser energy
• 13:00 - 13:30 and the rest of the operation is identical and just to give you one example what kind of images we can provide the red line shows the current cross section being imaged and uh we're actually imagining a 50 hertz frame rate limited by the laser reference with a single laser shot within uh tens of microseconds we got a 2d image so this has no motion artifact to worry about at all and this is one of the fastest imaging modalities out there for
• 13:30 - 14:00 biomedical imaging then we can by repeating laser pulse we scan our current current cross section and we get different cross-sectional images you can see different organs very well and drug discovery is the one major application of this technique previously using ion engine radiation for example x-ray ct if you want to monitor the same animal for tumor treatment over multiple x-ray ct sessions you have to worry about radiation dose
• 14:00 - 14:30 because the animal may die from the overdose of radiation before you get your results with our modality you don't have to worry about that so you get a non-net radiation entirely safe you can monitor a lot of functional information on top of your drug distribution now you can also think of other basic biological research to use this technology for as well we were excited about some animal data we scaled up our system immediately for human breast imaging uh you can see
• 14:30 - 15:00 this is a breast version the system is bigger of course to accommodate the breast and we have only slight deformation against the chest wall a chest wall this is a painless very different from actually mammography where you have to squeeze sideways very hard to get good contrast good resolution here there's only slight deformation for stability reasons and minimize the light penetration requirement again with a single laser shot you get a 2d image then you can get a 3d image
• 15:00 - 15:30 really fast this plot shows the imaging penetration versus the spatial resolution that can be provided using our technology and now this first version talks about at a very high end for deepest penetration and of course you have to scale your spatial resolution and from here you can already tell we're imaging whole bodies minimum organisms or human organs and you can scale downward all the way toward organelle level
• 15:30 - 16:00 even in the future toward protein level and this type of multi-scale imaging should find white applications in biomedicine because again i believe the future of biomedicine is multi-scale or omni scale and we have to relate information acquired across length scales and there's no second technology that allows us to do so and this is a representative image of a healthy volunteer
• 16:00 - 16:30 that's a nipple region a single laser shot gives you a 2d image which is acquired within 150 microseconds that's extremely fast as you can imagine this is limited by the acoustic transit time across the field of view within a single breath hold either 10 seconds or 15 seconds will acquire an entire 3d image of the breast we can see tiny blood vessels the smallest we can see is about a quarter of a millimeter in diameter without injecting any contrast agent for
• 16:30 - 17:00 mri to get an image like this oftentimes you have to inject heavy metal contrast aging like gadolinium currently we provide about four centimeter penetration and that's enough to accommodate most breasts and this is a cancer patient x-ray actually missed the tumor um ultrasound picked up the tumor then we label the tumor on the skin with a metal dot now this is a second mammogram with a
• 17:00 - 17:30 metal dot you see that dot right here using our technique we can see the tumor quite well as indicated here it seems to reflect this dense collection of blood vessels you can see a higher vessel density presumably associated with angiogenesis and hallmark of cancer as a bonus contrast because we can image it so fast we can track the motion of the 3d tissue and allow allowed us to get this strain map you can see this tumor area is stiffer
• 17:30 - 18:00 than the surrounding tissue so more contrasts give us more confidence about diagnosis again because we can image at 50hz frame rate we can imagine fast enough to see the heartbeat induced motion and we hear color code arteries versus veins you know we pick two pixels for example to track the dynamics you can see here pixel one has a much greater fluctuation because that's our due to
• 18:00 - 18:30 the artery you got a pulse coming in and that causes motion as you know the vein is after your capillary filtering and so that doesn't really generate much motion and we can image basically we can separate arteries and veins really well using this technique that's a form of functional imaging as well here is a recent study where we use our technique to monitor neoadjuvant chemotherapy for breast cancer patients on the left side we have our photo
• 18:30 - 19:00 acoustic imaging data on the right side is mri imaging with gadolinium contrast agent as we know gadolinium actually has some side effect to worry about you can see we even without injecting any contrast agent we can image more blood vessels than mri and we can see the progression of the treatment using our technique then we modify the system for human
• 19:00 - 19:30 brain imaging as we know brain science is probably the ultimate frontier in biomedicine we need more tools to understand how our brain functions you can see the system here the patient lines in supine position and we can image the brain using our modified system i'll skip some of the details about our system uh that you can see the laser comes down from here now this is all within the safety level
• 19:30 - 20:00 because we expand the laser beam within the nc safety limit you can imagine without causing any harm now this is a upgraded version of our system you can use it for breast or brain so i just came back for an example of breast image and this is a animation showing 3d breast image showing a lot of blood vessels imaged
• 20:00 - 20:30 using this system and we're ready to use the system for breast cancer imaging you can also use it for some animal imaging this is one quick example where we can see a dense collection of blood vessels we've all heard of obama's spring initiative one of the key goals is to understand how even a smile animal brain functions in order to understand you have to
• 20:30 - 21:00 understand in 3d volume ideally non-invasively how action potential these are electrical signals how they propagate how they connect and one way to see the action potential is through some voltage indicators that allows us to confirm convert this voltage signal into optical signal it's very hard to do that with any other form of imaging so optics has a unique role to play however standard optical medium does not allow you to see the whole brain at high
• 21:00 - 21:30 resolution our technology is ready for that application because we can see the entire brain on top of exogenous contrast i just talked about with some sort of conversion you can also see endogenous these are blood induced endogenous contrast that's intrinsic you don't have to inject any contrast agents so the movie actually showed you some of the examples fingers now this is a human brain
• 21:30 - 22:00 functional imaging we started with hemi craniotomy patients so their skull is temporarily removed but the scalp is intact and so they are safe to be imaged you can see the print functions you see this activation right here as the subject have fingers here we present results from subject 1
• 22:00 - 22:30 7t mri on the left functional packed on the right individual functions are flashed in starting with finger tapping lip puckering tongue tapping passive listening and silent word generation vessels for co-registration and then labeled one through four so one and two being superficial temporal artery three and four being cortical arteries the images are overlaid and then
• 22:30 - 23:00 individual functions are flashed in starting with finger tapping lip puckering tongue tapping passive listening and silent word generation
• 23:00 - 23:30 the vessels are again pointed out individually vessel one and two being branches of the superficial temporal artery three and four being cortical vessels the lateral perturbations allow for visual spatial appreciation of the co-registration images are then separated and native t1 masks are faded in again 70 mri on the left functional packed on the right individual functions are then flashed in
• 23:30 - 24:00 starting with finger tapping axial and coronal images through the area of interest are introduced below next lip puckering tongue tapping passive listening and silent word generation
• 24:00 - 24:30 all right so the mr data was acquired using the strongest fda allowed strength seven tesla and so we we actually i will present the table comparing the performances now of course so far we don't have skull the next challenge is to overcome the skull problem the skull is going to generate away from distortion so it's called aberration we have to
• 24:30 - 25:00 find a way to de-average to form a sharper image we've done x-vivo experiment already to show the feasibility and this is a x-ray ct image of x-vivo skull and that's a photograph and we put a canine brain inside the skull and we form the image like this so we can actually go through the skull and form an image which is reasonably sharp even without the aberration yet and we're working on the physics side to generate even sharper images
• 25:00 - 25:30 now a big question is why do we need to have a new modality and this is a table showing why photo acoustic tomography is important first of all we have richer contrast mechanisms function mri is mainly sensitive to deoxyhemoglobin and we're sensitive to both ioxia and deoxyhemoglobin it allows us to derive the oxygen saturation and the total concentration of hemoglobin response time is shorter you can see
• 25:30 - 26:00 here it's the shortest form is 6.1 seconds versus where both mri even at 7 tesla requires 7.8 seconds our signal background is low our signal linearity is high so it's more accurate for us to extract quantitative information our system is reasonably portable we can wheel this to the or or to the patient site bedside where mri is kind of very bulky
• 26:00 - 26:30 and it requires a big magnet so it's not compatible with many equipment and our platform is open we don't cause claustrophobic response this is a key problem in mri as well which has a closed bore our acoustic noise is low okay with mri that can generate a lot of noise our system cost is much lower than that of mri and again we don't need to use a strong magnet
• 26:30 - 27:00 so you can see a number of reasons why developing this technique is important now let me scale down to the microscopic world in 2005 we published this first paper on the first 3d photo acoustic microscope this is a key component here we start with this hollow beam we refocus into the tissue on the tissue surface we have this donut shaped illumination
• 27:00 - 27:30 so the core of the beam on the surface is dark we made it dark intentionally to minimize surface interference the ultrasound detection is confocal with light elimination to maximize the signal to noise ratio with a single shot you get a 1d image then we wrap our scan to get a 3d image now it jumped from roughly here to here now we get multiple millimeters penetration but now we get tens of
• 27:30 - 28:00 microns resolution and this is a representative image of a human palm you can see blood vessels again without injecting any contrast agent you can get a b-scan image showing you depth-resolved information that shows the skin layer structures now as we know that the primary cancer doesn't necessarily kill the patient
• 28:00 - 28:30 it's really the metastasis that's lethal detecting circulating tumor cells or cdcs is important but right now we don't have an immense tool to detect ctcs using photoacoustics we have a chance and we start with melanoma ctc because melanoma cells are labeled automatically with melanin and are highly visible to us we brought our photoacoustic system to the clinic we image this cancer patient in vivo
• 28:30 - 29:00 you can see here this bright dot is actually a circulating melanin-containing particle presumably a circulating tumor cell you can capture that several times with our field of view and this is another case where we see a doublet appearing multiple times consistently within the field of view and so this can find very very important potential applications in not just screening but also diagnosis prognosis
• 29:00 - 29:30 monitoring of chemotherapy you name it in fact that we can even follow with a stronger laser pulse just to kill the cdc on the spot and so we've done that in small animals already in but you know we haven't gained approval for human study yet you can see the potential there we can scale down further to get even better spatial resolution we call this optical resolution photo acoustic microscopy where we get single digit micron level
• 29:30 - 30:00 spatial resolution and this is the example this is our system working this as a demonstration and we operate fast enough to see individual regular cells in vivo and we can see how the cells split at this bifurcation point you can see it's not totally random and then we can also image the oxygen saturation of uh hemoglobin you know red blood cells are used to deliver
• 30:00 - 30:30 oxygen to different parts of the body how are they delivered and we can provide this tool to study at the most fundamental level single red blood cell level and this is actually human you know we're looking at our finger cuticle and we use color to label different cells if you single out one of the cells you can watch how the color varies as oxygen is released
• 30:30 - 31:00 the brain response we talked about seconds response right for both for both mri and photo acoustics can we do better than that and to understand to answer that question and we have to start from the small animal study and this is a response where we have a single impulse stimulation and we would watch the response of the spinal to that impulse and you can see a wave going down here
• 31:00 - 31:30 okay so this is real-time response sub second scale response so the innate response is actually very fast the question is how can we detect it with high enough sensitivity to get to that level to make the next generation technology if we look at the detailed response this is the fractional change of the signal versus time this is our impulse tens of milliseconds
• 31:30 - 32:00 in length you can see the onset time of oxygen saturation is only 130 milliseconds so this is water magnitude shorter than what the current technology can provide now if you look at the total concentration of hemoglobin measure signal for response it's roughly three times longer than the oxygen saturation response so we can optimize and pick the right parameter to detect the fastest response now you can imagine
• 32:00 - 32:30 this can provide a non-invasive tool non-invasive tool for brain computer interface to help people with paralysis and the speed is important for that kind of application pathology has been around for a long time and still the gold standard for disease diagnosis however we have to cut tissue from the
• 32:30 - 33:00 patient we have to go through a lengthy preparation procedure and cut it into a thin slice before we can examine the slide and give the patient the diagnostic result it's not real time it's far from real time can we do better and we're studying this label free without staining so there's a very rap head and eventually even in vivo we don't have to stain the tissue we don't have to cut the tissue to begin
• 33:00 - 33:30 with we acquired this photo acoustic microscopic image without staining we're using a dnra in the cell nuclei as the intrinsic contrast our pathologists can diagnose already using our slides they can tell apart the normal region from the tumor region to validate our result we compare with the standard hd stained histology results you can see how well they correlate
• 33:30 - 34:00 we're building the next generation for real-time imaging right now as we know that lumpectomy is not 100 successful in fact roughly one-third of the patients have to come back for a second surgery because we cannot remove all the cancer in one go now with our technique we can check the margin interoperatively so if the cancer has not been fully removed we can just come more
• 34:00 - 34:30 right there on the spot we can push the resolution even further to the next level this is the cell organelle level that's the structure within the cell just to give you one example with a very sharply focused lens you can get a resolution roughly 234 nanometers which is already fantastic but we can do even better by using nonlinear response we can push it down to 90
• 34:30 - 35:00 nanometer resolution and this is actually called monoscopy we're looking at a mitochondria you know at this level of resolution you can resolve some of the internal structures of mitochondria and that's a comparison between the two and as an ear micrograph you can see for comparison so i've talked about the multi-scale imaging capability of photo acoustic tomography
• 35:00 - 35:30 our lab had been worked working on the space xyz for decades then we said well let's not forget about the fourth dimension time the current imaging technology had a problem with with a temporal resolution the temporal resolution is not high enough to see light propagation as we know light propagates at the final speed the ultimate speed nothing can propagate faster than the speed of light we want to study light we want to
• 35:30 - 36:00 understand light better and so we studied this fast camera we call that compressed ultra fast photography our first generation allowed us to acquire 100 billion frames per second at this speed we can see a single light pulse in real time there's no repetition required just one pulse as it propagates we capture the motion this is one example where you can see the light poles propagating in some scatter medium and then gets re
• 36:00 - 36:30 reflected refracted and we can see two light pulses having a race here we see fluorescence with a single uh single light pulse excitation we look at it for fluorescence decay we made our camera two colored so it's a colored version and the green channel shows the excitation pulse the red channel shows the fluorescence decay the decay only lasts nanosecond scale but we empty so
• 36:30 - 37:00 fast and we slow down all the movies by 10 billion times to see the process you probably realize already you're watching the slowest uh slow motion movies right now right so quickly talk about how our technology works um how do you get that kind of imaging rate let's start with an object that has intensity i versus x y and time t we round this
• 37:00 - 37:30 intensity pattern to a diesel micromirror device and this is a device that's used in a computer projector for example we write this uh something called the pseudorandom code you got ones and zeros written on the micromirrors and so the pattern reflected from this dmd will uh will be modulated by that pattern that's the encoding process and then we round this beam to this
• 37:30 - 38:00 camera it's a it's called a streak camera unlike the standard operation of the street camera we actually open up the slit as widely as possible where standard operation use a very thin slit because the vertical dimension is reserved for temporal resolution and that cuts all the information on the vertical axis so you got no spatial information on the vertical axis we want both x and y information in a single movie
• 38:00 - 38:30 the next step is to uh now first we convert photons into electrons the electrons will be pulled backward then we apply this shearing voltage it's a voltage ramp this is very much like the oscilloscope and you apply this voltage you're going to shear you're vertically going to pull the electrons and so depending on the time arrival of the electrons you will shear the electrons to a different angle now the vertical dimension will carry the temporal
• 38:30 - 39:00 information it's a way of freezing the photons it allows us to get extremely high temporal resolution on the back end we will convert electrons back into photons and then we use a ccd or cmos camera to capture at a low rate so now you can temporarily integrate all the information without worrying about temporal resolution because we've already encoded the temporary information using the y axis so mathematically we can summarize
• 39:00 - 39:30 everything i described into a single equation you got intensity i the coding pattern c that's written down here by the dmd and there's a sharing operation from this street camera and then you got this temporal integration uh you have one exposure there's exposure time that's the integration operation then in the end you get an energy matrix e versus x y all we have to do is to invert this equation to invert this equation
• 39:30 - 40:00 to recover i because i is the image we're after or the movie we're after now you see a little problem here because e is two dimensional where i is three-dimensional so i has more unknowns then e provides measurements all right whenever you have more unknowns then you have measurements you cannot solve the equation set all right as our middle school teacher told us and they're still correct so there's
• 40:00 - 40:30 this modern version of math which is compressing when you use your smartphone to take a picture you can compress it when you take a video you can compress even harder yeah because adjacent frames are correlated there's redundancy right and so using that redundancy we can actually compress our our i so the actual number of unknowns is smaller than the number of measurements now the equations that can be solved and
• 40:30 - 41:00 allows us to get a movie like this there are many applications one of them is to study fundamental physics um you know that this is from nasa so when this airplane travels supersonically you will see this cone structure this is called the supersonic mod cone okay and it's also called a sonic boom okay there's optical counterpart but there's no tool that allows us to study this
• 41:00 - 41:30 there's something that's known in physics already in theory it's called cherenkov radiation and that where you've got a particle traveling at super luminal speed meaning you have a piece of material so your speed of light in the material is lower than the speed of light in vacuum and then the particle travels at a greater speed than the speed in the material then that's a supernormal situation we can mimic that situation in the lab and so we have this air tube
• 41:30 - 42:00 where the light pulse can travel at nearly the vacuum speed but you got material so the speed is lower and that's a super normal situation you can see this cone structure monochrome structure this is a super nominal optical micro so for the first time we can see this cone structure using our technique we know that our vision is sensitive to
• 42:00 - 42:30 intensity contrast the phase contrast is not readily readily visible to us and there's this known microscop microscopic technique called phase microscopy right we follow the same concept and we incorporate that into our technique and this is a phase contrast version of our technique and for demonstration purposes you can see this real time away from propagating acoustic waveform propagation you can imagine cells if you don't stain
• 42:30 - 43:00 cells the cells are not highly visible to us because there's mainly phase contrast but with this technique you can see cells by using the phase contrast information another application is chaos and this is a very important physics problem right so kiosks by definition is not repeatable so whatever behavior you observe is not repeatable between different runs and so that means you cannot use the
• 43:00 - 43:30 pump probe technique because next time you probe it using a different pulse you try to repeat the event it's not repeatable and so that's a fundamental problem for that field now we can actually do it with a single shot you don't need to have repetition and this is a cavity it's a chaotic cavity uh there are two different takes we're sending a very short light pulse they appear as a ball very much like a ping pong ball bouncing around well if you look at just one realization
• 43:30 - 44:00 you you really can't tell much difference but if you have two takes you overlay them you see that initially they travel together but if you wait long enough they'll split so this allows us to understand what's going on in this type of chaotic behavior much better and provide a tool and this is actually a controlled version we study transition from a non-chaotic cavity like this square box rectangular box is a non-chaotic version
• 44:00 - 44:30 and once you open up this gate this is a light speed gate you have a chaotic version and so that studies the transition during the flight time of this single light pulse we more recently advanced the speed to 70 trillion frames per second and so this is the current record world record we compare with the pump probe approach
• 44:30 - 45:00 the pump probe requires you to repeat the event and you fire a post you capture one frame only at a given delay time then you vary the delay time and you repeat the whole sequence you get another frame then you get many frames later on you piece them together to get a movie but again our system just have a one shot we get all the frames uh in one go and so this the new version is called cusp technique um this red dash line is
• 45:00 - 45:30 from our previous version and the new version is has a better temporal resolution so you can see faster temporal changes and the the dotted green line is from the pump probe approach you can see our data is much smoother than the pump probe approach because anytime you want to repeat the event you cannot guarantee it's highly repeatable your event is not highly repeatable your system is not highly repeatable then you got all these fluctuations going on
• 45:30 - 46:00 this is a chart showing the different versions of high speed imaging in terms of the number of frames or sequence stats per movie you can provide in the imaging speed you can see our cup family techniques are here it's the highest in terms of both number frames and the speed finally for more information please visit our live website
• 46:00 - 46:30 and we have a couple books that will give you more information i'm very grateful to caltech for building me a dream lab we have nanoscopic imaging at proximal end of this floor it is the one of the force overlap and at this land we image humans credit goes to lab members we have a lot of talented hard-working lab members we do have openings for graduate
• 46:30 - 47:00 students and postdocs we're mainly funded by nih and partially funded by nsf thank you very much thank you for that lee hong i especially love the little kid in the corner of the picture doing his power ranger thing who is that uh that's like that's the song of one of our postdocs it was very very cute i'm sure that he is on his way to uh to a career in science just like his parent
• 47:00 - 47:30 i hope so too so we're going to be taking your questions for dr wang uh you can put them in the q a box that you see at the bottom of your zoom screen and we will get to them get over to them and i just i have to say that what arrested me of so many things as i was listening to when i heard the subject whose brain was temporarily removed only the skull okay good all right that was that was still a bit of a shocker then so um yeah yeah there is a blood pressure on the brain pressure built in
• 47:30 - 48:00 uh due to some sort of trauma and so they have to release the pressure temporarily then they will put the implant back and so this is our window of our opportunity to study whether we can capture the functional information at all and it seemed to have worked extraordinarily well it worked really well yeah let's see what kind of questions we have here and it can get to uh for all of you from john who got his degree in 71 he says what advantage does photo acoustic have over the much
• 48:00 - 48:30 smaller scale technologies like electron microscopy right you know electron microscopy is for that tissue only you know in fact you will hear some experts they'll say well you go through a number of steps they say by the time we get the image the tissue is very very bad and so we imagine vivo we image live tissues you know so we you know when you take
• 48:30 - 49:00 the tissue out a lot of the physiological parameters are no longer accessible you know if you want to image the functional information you want to look at the oxygen saturation of hemoglobin for example right that's not available after you take the tissue out uh just to name a few and also for the operating room application just give you another example right and so you want an image right there on the spot you want to image ideally on the patient you know if you can have a pen a hand-hold probe you just say hey spot
• 49:00 - 49:30 check here is this cancerous or not if not you just remove more so you get instant feedback um even in a an immediate surgical setting exactly exactly and this is why electron microscopy is not used in medicine for that purpose you know because it takes very lengthy preparation and you know even the standard pathology will not give you that kind of information so looking perhaps a long ways out or
• 49:30 - 50:00 maybe not so far jeff who was graduated in 15 and peter who got his phd in 75 both want to know could photo acoustic technology become routine clinical use well so i mentioned very briefly that one of the companies received fda approval for human breast imaging early this year and this is going to uh accelerate fda approvals of other techniques i believe right so they really paved the way it took us a long time to get here
• 50:00 - 50:30 the field has worked really hard to get this far so it's a very new technique it involves two forms of energy so the fda has to really steady make sure we're safe from both the optical perspective and ultrasound perspective i mentioned it very briefly that light is very safe as long as you broaden the beam anti-regulates energy per area so if your energy within uh is within 20 millijoules per centimeter squared and you're operating
• 50:30 - 51:00 in the visible spectral range you're safe and so the sound produced is orders managing weaker than the sound was sent in for routine ultrasound imaging so this technology is entirely safe and fda had no problem uh approving this technique in the end the other proof we have to show is efficacy you know is this technology um adequate right is is this good enough for making disease diagnosis
• 51:00 - 51:30 so obviously this company i mentioned has done a great job they have clinical data published maybe about a year or two ago to show they've done a large study thousands of human subjects you can see this is very laborious to go through fda approval in the meantime some animal research products are widely available because for that you don't have to go through fda approval and they're benefiting a lot of research labs the devices are being sold around the
• 51:30 - 52:00 world mark has had got his doctorate in 90 and wants to know living things must be full of acoustic noise i think we heard a little bit of it in the video you provided he said the ultrasonic component of that just not an issue what portion of sound frequencies do you use it's a great question we use a very high frequency ultrasound uh depending on the depth for brain imaging we prefer to use about megahertz for
• 52:00 - 52:30 breast imaging we prefer to use about two megahertz for microscopic imaging we use tens of megahertz for example 50 megahertz or even 100 megahertz and so the acoustic noise is very low and also as you know that that's like background random noise uh they're asynchronous with our signal we have time zero the minute we fire laser pulse for excitation that's our time zero then we detect over a very short period of time you know for
• 52:30 - 53:00 the breast scale we detect over maybe uh 200 microseconds you know roughly about 150 microseconds for smiling imaging we detect down the order of tens of microseconds so you got a window to reject all the noise you got synchronization you know what time zero is to reject the noise and so that's why we get high quality images so you anticipated that and built it into the process
• 53:00 - 53:30 we can claim so but it was never a problem [Laughter] oh go ahead um jan who was graduated in 63 wants to know could acoustic imaging be coupled with proton beams that have deeper tissue penetration another great question in fact has been done um you know not too far from where we are there's loma linda they actually have one of the first proton beam systems decades ago i actually wanted to work with them i never went anywhere all right so we
• 53:30 - 54:00 said you fire proton beam one of the open question is what kind of dose are we delivering so you got proton beam that generates some acoustic signal as well when energy is deposited and they got this frag peak right that's why the proton beam is so uh so attractive you know you minimize uh damage along the way and then you got a huge peak where you deposit energy where the target is uh they say well how is the dose adequate right so if we have a non-invasive tool
• 54:00 - 54:30 to monitor as you deliver protons then at some point it's everybody we got enough dose so a few years ago another group in europe decided to do what they actually demonstrated this is feasible so you're exactly right has been done now yeah um esperanza who got her phd in 2015 wonders and this goes to your question about fda approval too does this technology cause any tissue damage
• 54:30 - 55:00 so i you know sort of answered that question already uh there's none whatsoever but you need to conform to the nc611 because laser can be used for surgery as well so if you intentionally want to do surgery you can focus laser light so it's all about those controls just like anything else like ultrasound right ultrasound can be used for imaging but ultrasound can also be used for therapy you know we have high food right where we can high intensity focus ultrasound where we can
• 55:00 - 55:30 burn tissue literally inside tissue for therapeutic purposes x-ray the same thing you got diagnostic those then you got therapeutic dose you name it and so within our diagnostic domain this technology is very very safe um michael who was graduated in 93 said when imaging traveling light pulses in in a side view is it possible to image pulses in a vacuum or is it
• 55:30 - 56:00 only possible to image these pulses when there's some transverse scattering from the ambient medium yeah you essentially answer your own question um you're right so in vacuum you can only receive photons head on so when the photons reach your detector you can you can read the photons now the demo i showed we have some scattering materials like a fog uh you know so when your photon your light beam travels and you
• 56:00 - 56:30 scatter some photons sideways towards your camera so you have to receive photons before you can you can image or film i think your results with breast cancer your findings have been very arresting because there are a lot of questions about that one from daniel is with the breast imaging can you differentiate invasive breast cancer from ductal carcinoma in situ and can it differentiate histological subtypes of the invasive breast cancers that's definitely our
• 56:30 - 57:00 goal right now we don't have enough data for that claim because we can image not just the blood vessels as we've demonstrated we can differentiate arteries from veins we can look at oxygen saturation of hemoglobin as we know that there are two major hallmarks of cancer right one is uh hyper metabolism right cancers grow they're aggressive they consume a lot of oxygen they deplete the oxygen in the blood vessels and so we can see the oxygen saturation
• 57:00 - 57:30 to quantify the level of hyper metabolism the angiogenesis is another hallmark we can detect vessel density and look at exactly what's going on there then on top of it as we've shown we can image some acoustic properties and mechanical properties like the stiffness of the tumor and then we because we're internally non-invasive we have another dimension which is time the patients can come back in the image as frequently as possible and it's fast and so you can just monitor the
• 57:30 - 58:00 progression and see how those parameters vary over time the hope is that at some point we have enough data to answer your question better um let's see um we have joe who's wondering whether you tried detecting muscle trigger points with this uh not exactly not intentionally at least you know so as long as there's any optical contrast you know we we've done all sorts of
• 58:00 - 58:30 stimulations you know like whisker stimulation electrical stimulation you know finger tapping humans you know town tapping all of those right so if you have any um you know basically hemoglobin type of response we can potentially track that and we spoke about the uh the patient we saw in the video with the skull removed for access so jeff um who got his masters in 81 says given
• 58:30 - 59:00 that the skull is one or two centimeters thick as as you showed does that interfere with the imaging and is there at some point a way to not where the skull will no longer be an impediment that it is right now so again a very good question right so um as i've shown in the ex vivo study um ultrasound at the right frequency can actually go through the bone quite well and we can form an image even without
• 59:00 - 59:30 um the operation with what you know we call this physical process of sharpening the image the aberration so the skull causes two problems man one is it functions like a bad lens you know optically if you have a bad lens you're going to see a blurry image and the other is attenuation so the sound signal will attenuate as it goes through so the best frequency to use is roughly one megahertz in fact in the lab we found that the optimal is about point eight megahertz
• 59:30 - 60:00 now that kind of frequency you can still provide roughly one to two millimeter spatial resolution which is adequate for a lot of applications including functional imaging and so the xvivo study has shown that we can form images we can deal with the attenuation and our current system has shown the visibility of in vivo study and to push to the next level with skull we actually receive the grant to do that we have to make sure our detection can have an enhanced sensitivity
• 60:00 - 60:30 and that system that was used for individual human brain imaging was sub-optimal for brain imaging it was actually built for breast imaging so the frequency was wrong the system design was not optimal and so with our next generation which is being built and our hope is that to push to the next level of sensitivity to see functional brain imaging and there's also data i didn't have a chance to show is in full normal human subject
• 60:30 - 61:00 we can image transcranially we can actually see the brain cortex already but we are not sensitive enough yet to see function so we can just see the structure but not function so you can use the same platform that you would use say for breast tissue but you just have to tweak it you have to tweak it you have to optimize that for example the ultrasound transducer frequency has to be optimized the geometry has certain ways of optimization and so we have several ways of optimizing for human brain imaging that's working progress thomas of the
• 61:00 - 61:30 class of 60 wants to know whether you can exploit polarization properties for this technology another great question um so actually we've done that um so light absorption sometimes has die attenuation something called die attenuation that's polarization dependent attenuation if you have a linear polarization you rotate at 90 degrees you measure one way then you rotate 90 degrees you may get a different measurement and so that we can do that
• 61:30 - 62:00 but polarization will scramble as you propagate too deep into the tissue so this only works roughly within a few millimeters um eric of the class of 67 wants to know beyond morphology pathologists use special stains we know to identify features that are indicative of tissue origin or derivation and tumors is there anything you can do comparable to that using your technology flagging those things yeah again i um i didn't have time to show those
• 62:00 - 62:30 images we've done that work it's that field is called molecular imaging and that's basically a labeled version um so you inject some sort of contrast agent and these are safer versions than histology right histology doesn't care about the safety they just want to stay in certain features um so you can label you can target different hallmarks you know these are called biomarkers you know then uh including the the version i mentioned very quickly which is a
• 62:30 - 63:00 voltage indicator that's actually a way of converting your signal from electrical signal into optical signal and you can do something very similar using dye dye molecules nanoparticles right then you can target maybe angiogenesis hallmarks alpha we built a three integrand for example right so that has been demonstrated in small animals unfortunately none of those are approved for human use yeah so you can start with unlabeled versions fda has a house grandfather
• 63:00 - 63:30 grandfathered a few dyes like icg uh you know with the endocyting green that that's allowable for human use right so we actually plan to do that as well it's entirely possible um ella who got her masters in 78 says breast images show a lot of blood vessels is it possible to selectively mask out blood vessels maybe see more clearly the underlying fibro breast tissue that's there so
• 63:30 - 64:00 we can potentially tune the laser wavelength that's another knob we can we can tweak and so different absorbers have different peak absorption wavelengths and so you can tune for example you know you can tune to roughly 1200 nanometers to see lipids better right and you can to another wavelength you see the water content better and that's a possibility that's sort of a way of masking certain contrast
• 64:00 - 64:30 you suppress certain contrast this is the last question it's very quick and it's very touching coming from deeper who's an incoming freshman saying will your lab consider undergraduates too if someone is very interested yes uh we we third we have um openings uh not many a few for undergrads as well yeah i'm sure he'll be glad to hear it depro thank you for asking thank you all and thank you so much dr wang for enlightening us and giving us a literal window into this advantage this breakthrough
• 64:30 - 65:00 that may become standard a standard tool of science in a couple of decades time thank you so much we really thank you pat thank you all