 Good morning everybody. So today we will continue with the summer school and the topics of today is kind of trying to acquire data that we will need for our modeling. So we will try to focus today on which are the techniques and what are the ways in order to acquire the information that we need to do both validation or personalization of our flow models and We start first with trying to talk about the gold standard in some ways because what is very important is when we do modeling that we need to be able to validate whatever we have and The best validation currently is still doing like optical methods with Particle tracking and it's my pleasure to have Pilar Arroyo here Which is working in the University of Zaragoza and she's a physicist and leading a lab on Optical techniques in order to do measurements, especially of complex flows And these are the things that if we use complex geometries with complex flows We need to get like real measurement of as close as possible to reality so that we can Validate our model. So it's my pleasure Pilar. Thank you for coming Almost 35 years ago and all I have learned about vascular flows is because of the Interaction they have with groups that we're doing mostly simulation So they were doing simulation. They wanted to have some data so they came to us For obtaining this kind of data So because you are more interested on vascular flow that on the optical techniques I will start just with Summary of the kinds of flows we have been applying our techniques So the first one The first one we started with the aneurysms This is a medical image of what an aneurysm on the cell on the an intracranial aneurysm look like and It's like a balloon is when the the arteries Becomes weak and it can start growing and growing and growing and finally exploding so We got three images clinical image of real Aneurysm realistic from a real patient and we've built enough in glass We have a glass blower at the University of Zaragoza, which is was quite good So this is one example of this We also did some models with silicone which are supposed to be flexible and they really were flexible This was built by the company in Switzerland that are very good with this kind of modeling Then later on we work on this kind of problem This is related with trombosis formation and the problem that trombosis Gives when it reaches the heart or whatever So we wanted there was a group that wanted to simulate How the flow was going to be like on a vena cava when they put Antitrombosis filter that was to go to catch up the trombosis So this is a commercial filter simple one and This is a vessel model Where we put the filters in And I will show you some examples on results on this application Later on someone gave us a bulk silicone model of a carotid v-frucation So this v-frucation can also be patient specifically done because it's done with 3D techniques They excavate this model inside the silicone and Later on also we were we thought about getting data inside real vessels Which are not transparent all all of these models are transparent because our technique optical techniques They need to have optical access to the flow So we thought that maybe with an endoscope and we try to use Commercial endoscope we could reach inside the vessel if I have time maybe I will show you something about this This is the last part of my talk okay, so because the idea is to Give you information on what kind we get and what you need for simulations. I have a wrote down these these things here so Maybe you know already when you do simulation. What do you need? Well, you are going to need information on the geometry of the model 3d geometry because simulation always done 3d Nobody does 2d. It's always 3d. The computers are very powerful and the techniques are very powerful And also what I've learned is that what you need is of course flow rates this I can understand But also pressures and the input and at the output This is what you need for the simulation and this is what we are trying to measure on our experiments And of course liquid properties that usually they are about the same as the blood Although the blood is a non-Newtonian liquid and usually This part is not model. Usually it's Newtonian. There is little differences So what are your simulation variables? I guess usually is the velocity field and the pressure I'm not measuring pressure, but I can measure the velocity field and What are the measurements? They need to be non-invasive thing because I don't want to change the flow when I try to measure it If they are a whole field means at least a plane is better that if they are only a single point and Also, if you have high spatial and temporal resolution, that is much better because you can follow better how the flow changes with time So from my point of view the main characteristics of the vascular flows are they are liquid flows Which means that the velocity are lower than when you compare with air flows Which is better for experimental techniques. It's easier to take data They are confined because you see you have this shape or you have a cylindrical vessel It means the optical aperture is not big and also because the geometry is complex even a Cylindrical vessel is complex geometry regarding illumination and recording Makes imagine difficult, but we can remove this complexity by placing the model inside a rectangular box Fill with liquid with the same reflection in this of our model. This helps a lot This is what we have been doing usually and one good thing is that the the flows usually are either steady or periodic flows Which means that it's easier to get 3d data because if they are periodic you can take different planes for example always on the same point of the Waveform always so you can reconstruct a 3d image of or model with 2d data 2d means 2 dimensional, okay So now If we go back to what I like most is optical techniques for whole field measurements So in the group I'm working with we have been using these techniques as I tell you for 35 years I started really with this kind of measurements which is velocity in a flow in a plane When I was doing my PhD. So what have been God what have we been measuring with this technique we have been measuring displacement This can be done either on a fluid or on a solid for example on the vessel wall Displacement on the fluid means velocity When it's one plane we get the in plane velocities If we use a different technique we get only one component which is the out-of-plane component So we combine the two of them to get the three components of the velocity if it is not a Fluid but it has all a solid we will get deformation in plane deformation or out-of-plane deformation So well, I'm going to be talking most about this. We can also measure shape Okay, we can also measure surface corrosion In the range of micrometers when a surface is being is losing material because of the Agent that is being corroding it we can measure this we can measure refracton in this When something is being changed to liquids are mixing and we can also measure particle sizing. Okay So the basic features of our optical techniques for fluid velocimetry are drawn in here So this is the fluid area This is a illumination beam and Then we record this The fluid needs to have something that scatter the light that is being used for illumination This means it can be either natural scatters that are for example in the blow the Cells that are moving with the blood could be good enough if not for in-vitro experiments We usually put particles in there micro size particles and Because the sensors can only detect intensity. This is a photograph. This is a photography This is the first the very first technique This is well known in fluid velocity for a long long long time at the beginning of the years Maybe 100 years ago. They were using what is called particle tracking They were in the identifying particles following then on after the other and velocity is always measured as the ratio of displacement divided by time interval okay 35 years ago is when they started to do Statistical analysis correlation analysis. You don't need to identify particles. You can have more Seeders and then you can have more resolution higher spatial resolution And this is what is called particle image velocity, which I will talk to you in the next slide and which is now Very very very much used in every experimental fluid mechanics laboratory. There are commercials software and commercial equipment but The light that reaches the sensor which we will call the object in the light that scattered by the particles has information only No intensity, but also on what is called face The face of the wave. This is a wave coming and the face is related with the Path the light has been traveled to if the path is longer the face is bigger It's shorter the page is shorter, but there is no sensor that can record faces They only can record intensity. So how can we do the face recording then what we do is we mixed the object being with a beam that comes also from the same source and This face Differences from one point to the other change into intensity differences. This is called interferometry. So this is a Recording where the intensity here is directly related with the face difference Okay, the the only requirement is that this reference being has to come for from the same laser That the illumination of it is coming from because even if you have two lasers that are Identical one and the other they will never never never interfere because the face of the waves They are producing are never never the same So this has to come from the same laser, but this would need just a very small part of the beam So it's no problem. The only problem is that if you use for example white light The coherence led of white light is maybe an anometer. So if it is an anometer just exaggerating The path from one beam in the path from the other has to be exactly the same So the coherence ladies and I will talk more later on on these techniques Okay So as I told you we need to take two recordings at a suitable time interval that we can select with our cameras And with laser so that the displacement of the particles are what we want Either a few pixels or nanometers or whatever whatever the technique we are using on and as a rule The special time has to be a tenth of the time interval because otherwise if it is too long the particle would move and Instead of having a very nice and fixed position will be like a straight and this will not help Increasing the special time will not help in those circumstances so the intensity and phase change related to the displacement and We have high temporal resolution because it's fixed by the time interval and the recording speed Most of what I'm showing here has been used with a high speed Assistant, which means laser firing at 10,000 pulses 10,000 pulses per second Camera going at 1000 frames per second this kind of techniques. They are CMOS cameras And it's high spatial resolution the highest possible solution depends on the camera You will have a 2,000 pixel camera This is our resolution. It will go more spatial resolution. We will look at smaller area We won't even don't need such a big resolution. We look at a big area. So changing the area We're looking at we can change the spatial resolution Okay So this is the first one and then I'm going to do this Okay So the basic technique is when we use photography is the simplest one is this one and It's called particle image velocity as I told you When we use this on solid mechanics it used to be called a speckle pattern photography or digital image correlation Speckle means because when we illuminate with a laser anything We never see something which is smooth. It always has like dots Speckles okay, so in in solid mechanics is called a speckle when we use a laser and it's called digital correlation We use white light and put some paint on the surface, which is another way of doing it So for this kind of thing laser is not required We can use any kind of light But lasers are good because you can put more power in a narrow light sheet usually and usually we use a light sheet Okay, and This is an example Just for show these are these are not the real PAB images, okay, this is an image taken with the laser firing at something at about one thousand percent per second in the camera at firing at 50 frames per second so we have here a multi-spose image So we see really very well how the flow is changing. We will change the flow rate by hand So it's a it's a non-constant Flow rate so what are how the real image look like the real image look like this You see also one is pushing on frame the next in the second do you see the change in velocity Okay, so because we see which one is the first image and which one is the second one We can know not only the velocity of also the sense the direction if it goes going Right to left or left to right So how do you analyze this? Okay This is is using a static is a function with this cross cross correlation function. This is the first image This is the second image we select Area which usually which usually is 32 by 32 pixels. So this is our spatial resolution And the next window usually move half of this. This is the standard on the on this technique and This function give us a peak and the position of these peaks tell us how much the particles in this area has moved When we have changed the time in the second recording and you see there is one peak So it means they can move zero and we can record it We can they can move a half of a pixel and we can know it So the lower speed the lower displacement were measured are very low in the high usually 10 pixels or so We don't want the particles to move more than that But this cross correlation function numerically is not calculated as the cross correlation function It's calculated in a version where there are two Fourier transform that are being done. Okay, and this is being shown here This is an image from the wool good old times when I was doing my PhD no digital medium only Film so I have here this and let me see you see this Do you see something? Okay, is this Because when you illuminate something with a laser beam and look in the infinite you get the Fourier transfer of what is here If you put a lens on the focal plane of the lens you will see this Fourier transfer So this is the Fourier transfer if I go in the center We see let's see. They are horizontal so particles are moving vertically then I turn around and change Okay, so yes this the spacing and the orientation give me the modulus of the velocity in the direction and On the old times I was using something for measuring this with degrees and then measuring the distance with a ruler But later on you could do this on the computer Okay, so this really is the cross correlation function of this image with itself Which is main auto correlation function So when you have all this portion in the same recording because we don't have you don't have a camera fast enough But you have a laser fast enough This is what you get you get three peaks and the problem is that when the displacement is smaller This peaks mixed with this one and then You have errors and also, you know, which one of the two peaks is the direction of your displacement But this is the way it's working and what is the output? The output is always this This is the output of a software from LaVision. It's one of the main companies on this area And this is this plot here is Two components velocity map So these are vectors showing the direction of the velocity in each point in each 32 by 32 pixel region And the color usually is the modulus of the velocity or whatever you want to you want to do the horizontal component The vertical component you can plot it as you wish So the only thing you have to take into account is that this Vector is not fixed by this it's fixed by this it's always the projection or you Velocity vector in the plane perpendicular to the observation direction Okay In this direction so in this point it doesn't matter if there is motion here, but if we have motion here We have some errors on the outside. So this is being used For Measuring three components. How can we measure three components with this kind of technique? Like with the eyes it would take images from that plane from two different directions Then we'll have different projections and from two projections. We can get the three components This is called a stereo PAV. It's also a commercial technique and is Used very often in gaseous flows in liquids is a bit more complicated because the imaging through a liquid has some problems But you could do that However, because this is difficult in fluids, we have proposed a technique which I will explain you later on Where we can measure this third component with the same setup without having to look from two different directions, okay? so Just to not be too boring about optical techniques just I'm going now Showing to you an example a vascular flow sample Which is this one? The Venacaba filter. All we did with this in this application was PIV, normal PIV So this is a human body as you can see this is where they place the anti thrombus filter This is where the thrombus used to be formed on the legs and when they go up and go to the hurt and so on You have problems so they put it in here They put it for a while and Then they remove it and luckily they will have removed the thrombus you have there on that time so How did we Model this on the lab. We bought that piston pump where you can put different kind of flow flows physiological flows or steady flows and we will all this see fit And here is our model The Venacaba. This is all for measuring the pressure in and out of the pump because it has certain size tubes here And this is different tubes here and we measure also pressure here Inside and outside When the flow is steady, no problem. Everything is the same When the flow is periodic Changing with time like in a femoral or a carotid or whatever one of them then there can be changes on pressure both in shape and in position In in this kind of flows we didn't know that but I know fluid mechanics people later on I told to them and they already knew that for us It was a surprise to find this kind of So the the model the model for the Vena is just a silicon tube with this length and This diameter is not really really like a Venacaba. I've been told that the Venacaba is much thinner Not two millimeters in thickness. So when the flow goes to the Venacaba it spans and collapse much more But this was good enough for the people working on simulation So they what the model was this one even if this was was not really a Venacaba These are the pressure sensors we were using so we have to open up this and place them in here And these are the flow rate. This is based on ultrasonics So we could put it in and out without having to change the circuit and These are the flow rate of my pump and these are one of the waveforms we were using Okay, and as a fluid, this is the mixture we were using which for really for us was important That is similar to blood but also the reflection in this is quite similar to the silicon. So we didn't have Reflection in these changes while looking at it. So this is important Okay And this is a photo of the setup This is the camera with the tube in here The laser is coming from the top by using some optical arms that are being sold by the companies This is how the flow rate sensor looked like and the camera is here We also have another camera on there. We can use both and we can do one technique with the other one And the other technique with the other one. Okay This you can see here the Commercial leg and this is the model of our trombosis with it. So here you can see how the This is for a steady flow, constant flow, how it looked like. This is for visualization Okay, for measuring we have only one image in each frame And this we first check everything by measuring the flow in this area where there is not perturbed distort by the By this filter and we got the typical Poisson flow and we compare the flow rate We could calculate by interracial in this area with the flow rate that we measure with the pump So the agreement was quite good. So we were quite happy that the the Flow was working as expected That was the first test and here I show you the resolution. This is what corresponds to 32 pixels with this magnification Maybe you can see here that it looks like there is one, two, three, four, five different areas Okay, and this is because of course we have measured the whole area at the same time But then we will lose a lot of resolution in this area. So the camera is 2000 by 2000 pixel. So we want to look at an area which is almost square So we have to put together the different areas and for that we were using So an algorithm for edge detection this corresponds to this area and then just Have an overlap area from one image to the next one and we're using that to overlap We were moving the camera the we were moving the box this box in here Just shifting maybe 20 millimeters. We'll have a ruler there We move 20 millimeters again in the next region then the next one and then overlap all of them And I think the overlapping was quite good Okay Then these are The vector fields we can get out from those data and just shown Plots at different positions to see how they look like Okay, you can see of course after this this bit of the tip of the filter the flow Separates in two and we have this we can calculate the flow rate in each of these lines By interred in this area and we could measure that it was okay. Everything was was in good agreement and in here well here Here we have we have have been plotting in colors the vertical velocity Without the filter vertical velocity was null everything was horizontal But because of the filter here was going down and then up what you can see here on the on the vectors Just the orientation here is plot as a ISO contours for this component of the velocity and then In this paper Marina Nicolás that is working with you now as I believe She was doing the PhD on the simulation and she was using our data for comparison So this was what we measure that was she simulated and this is the comparison between Simulation and that Okay, we also measure with the trombus and They did simulation with the trombus, but what happened is that? Can you see something different on the filter from the previous image to now? Do they look the same filter or not? Does it look the same? Or not What do you think is it the same or not? Not what do you see? Yes. No, but this is the trombus. What about the legs? Are they the same? It's not really torn We always try to put this in a certain position because this is very important for simulation for comparison And we try to put it vertically, but what happens is that one of the times we have to remove The vessel put the filter put it back again So none of that's the person that was working with it remove it very quickly and the filter was squeezed and one is squeezed is Impossible impossible to put it on the suck shape as it was before so it doesn't matter So this is probably why they didn't compare because it was a bit difficult then Okay, and then we also did measurements with one of these waveforms femoral So you can see how it looked like and you can see that there is flow, which is the reverse negative and positive which was a surprise for us also and Also with the let's see One moment What happened? This happens. I have to Okay, well, well this is start you can put this on and Pass it around because I will try to show you something as look at the bright dot and Try to have a bright dot somewhere in here And then I will explain what you see is the right out there. Yes Pass it around Well, where are we now? This one this one this one we put it in here just in case Okay Are you finished? No, pass it around Okay, but I want to do this. No, I want to go where I go Where I am Okay So this was the flow with us. So you see more interesting than that is in it looking at the dot Well, well while you keep talking at the looking at these I can Go to the next application Just telling you first what we did with PAB and then I will explain you what you see with a With this thing we are looking at Okay, so the first application on the biological flows that we started with this one the analysis And these are the three patient patient specific models we got We had the images taken at the clinic with the MI that you all know and you like and We gave the 3d representation that dr. Franti was doing on the computer They could they there could be turned around and we gave this to the glassblower He did these three models the one on the bottom is three and a half centimeters It was in the head of someone. I don't know it was alive or not But it was in the head of someone the one in the middle is about one centimeter and the one above It was our half a centimeter the one above we never could measure it because there was a lot of Bubbles on the glass around it's not because of the size is because you all these bubbles So we have mostly studied the one below and the one above Okay, and this is the setup similar to the one before Okay, and these are the results this one only for show this is your visualization You can see the vortex that is being produced in here. It looks very much Similar to the shape of the aneurysm because the aneurysm shape is being Induced by the flow that is growing inside. So it means that the glassblower did a very good job Where are the bubbles still there? Okay, and This is this is the velocity field and this is the velocity field in this case for a plane that was in this direction Okay Because the the fluid was entering in here This this model is the one that was compared with simulations also Okay, so now we come back to this part, which is the optics again Okay, so as I told you already if I mix the object being with this reference being then my technique is not Photography anymore even though the sensor is the same is interferometry and interferometry Measure face differences only measure one data face. So it gives you only one information one component Okay, but the component Depends on the illumination direction and on the observation direction as I will show you soon enough, okay So in the old times when digital was not the usual way The technique was called holographic interferometry was made with film like the photographies and there are very nice images of holographic interferometry Then with the digital means started to be called digital speckled pattern interferometry and The images from this technique were just simply we record two Images which are called speckle grams grams because there is an interference between the reference and the object Just two of them we subtract them and makes the modulus which is the difference and this could be done even Analogically with electronics and the output was something similar to this. It was not the solid is I Don't know what happened with the previous one. It was this not similar to this. Okay, and Here just the color tells you the face I Will tell more and then they started changing Well, if I place this reference being in a certain position Then then I can do something which is called a spatial phase shifting and with a spatial phase shifting I can have something a bit better okay, and Then They realize that they could get not only the face but also the intensity so they could let the particle images With a little bit more noise that with with particle image velocity So then they call it Holography, so this is a name we call to this technique because we were the one that proposed this digital Speckle pattern interferometry for fluid velocity. It has been used for solids Deformation but never for fluid velocity. Okay Where are the goggles? Where are they are still there? Okay? So now is the video So the the image we get with interferometry They always go have this speckle noise or we have to filter it out Which is this and then here is just an example. Oh, no here a little video of Five horses images Corresponding to a constant flow, which was not really constant Constant fully constant because it depends on the pump and one thing I see one thing I remember I haven't told you on the photography technique Maybe you saw here that it says image space On the photography techniques on particle image velocity the measuring ruler is the speckles on your camera Then with magnification you go to object space So this is why I was telling you always 32 by 32 pixels displacement of 10 pixels So in interferometry the ruler is the wavelength of your light So it's always object space and the displacement is or the order on the wavelength So we are talking about Factor between 10 and 100 in sensitivity between interferometry and photography But in fluids this is not a problem because we have two time intervals one for interferometry and another one for PIV so we can measure the three components with different different time intervals, okay What are our goals or they are still Let's keep going, okay So this is the this is the example of this setup the fluid here This is PIV and then we put this using here the lights it is going from left to right This is the illumination direction and then this is the observation there So the velocity component you are measuring is this this is called the sensitivity vector is the bisector of illumination and observation From PIV we measure the components in this plane because this is perpendicular to the observation direction So they are not three perpendicular components But we can obtain the three of them by combining the information provided by the two techniques And this is an example of what it looks like an image blown out a small area and this is a fluid Okay, so you see this image. This is the this is It's called a speckle ground, but the image without the reference we will look similar So you say you can say well, this is a bright image a bright particle another bright particle another bright particle These are smaller particles. There is some kind of pattern like an interference So this like interferences due to the reference beam, but you see they are all circular So this is called a speckle ground because the interference of two objects two of two beams of two waves Okay, and just by subtracting by making the difference, which is the absolute value of the Substruction, this is what we get here and this is the equation that corresponds to this image There is a relation between intensity and phase difference But the problem is that there are other pattern here I not is the intensity of the beam at that point it can change So it's very difficult from here to obtain quantitative information Okay, however Where are the goggles? Then you have to pass them to part. Okay, so however what happened if we go and place this in a little bit different Position you see there is an angle between the object and the reference beam Can you see the difference from here to here? What do you see here? How do you how will you describe this image? How will you describe it? Not yet, but more specific No, because the focus is always circular the left The form in which direction? They are not circular. There are lines, little lines. Which direction to follow these lines? Do you see the lines? Do you see they are the lines are mainly like that? 45 degrees so what it means is that by doing this in each of these dots Which we can think of them as speckles The speckles has always the same phase. I have introduced a modulation This is what is called a special phase 15. Okay, so this modulation Allows you to say well I can get the phase of each of these dots by comparing the intensity of three pixels one next to the other because we have Here a modulation introduced by the angle So this is why it was called a special phase shifting and then we get this kind of images Which is what I really showed you before in this case is not very clear here. You have black grey white Black grey white black grey white. So you have the sign because here you will have Black and white and you don't know what is increasing or decreasing So with this one you can know if the phase is increasing or decreasing and you have once you filter this out Direct relation between the grey level and the phase. Okay, and then This is what you saw Isn't it? Yeah, I can show you also these ones This is what is being used on Pointers When you have money dollars dollars dollars This one is with the An opening And do you see apart from the central one there is a one top button left and right. Do you see? Okay So this is what is called Let's list Fourier transfer of something because when I put the laser beam And look at the infinite. This is the Fourier transfer of what is here when you look with your eye Through this Again, you see the Fourier transfer on your eye because the lens on your eye is almost at the focal length So these are all of them this one also that Fourier transfer holograms. So a Fourier transfer holograms Just requires that the reference beam is divergent Which is good for digital means because when it's divided like the object being the angles are Smaller, so we don't need a high a very high resolution spatial resolution We don't have when we compare with the film. Okay, so this is a good option And when you do the transfer transfer what you get is whatever it is on the point on The focus of the reference beam, okay? So with the with the image I was showing if you do the Fourier transfer you get this so what is this? this is The aperture that we have on the lens and This is because we have placed the origin of the reference beam at the same distance as the aperture of the lens and in a hologram and This is and this is a Fourier a Lensless Fourier transfer hologram of the lens aperture So we have in the middle the beam that is Direct one is not being the form and here one of these is the real image of the aperture and one of them is the Virtual image of the aperture the advantage is the face On one of the aperture is just the opposite of the other one So if we select one of these apertures and take the inverse Fourier transfer will go back to our Sensor plane and we get there the face and the intensity will get both of them Okay, and if we compare subtract two faces we get this Which are image usually the cleaner Yeah, that just stuck in the spec around because you see we have removed all these things that concern everywhere You don't know when you do this interference There is anything on your field of view that can't in that can be included there that can influence your spec around Okay, and this is one you filter and unwrap because here the face You only know it's zero to five But we don't know if it is zero to two pi or two pi to four pi There is this unknown that you can only Calculate it by unwrapping unwrapping means that I see the faces increasing So I put zero here when I go to two pi the next one has to be a bit bigger than two pi So you will go for pie three pie and so on We'll plot it wrapped because it's easier to see it. Okay, so this is what We can do in this case and this is a typical hologram reconstruction here is the focus of the reference being here is the aperture They are on the same plane. This is the object. So this is this this was one Our proposal for using a fluid velocity. So here is the fluid plane So we are doing an image plane hologram of the object This is why it's called digital image plane holography because on the sensor Which is the hologram we have the image of the object But it's at the same time a laser Fourier transfer of the aperture, which is what allows us to Discriminate between the real and the virtual images because otherwise they will mix together. Okay So now I As going to show you an example of the wood all times. This is my PhD 35 years ago Also, this is the image. I was showing you the the fringes here And this is a relevant convection cell We heat up the bottom we keep the top at a fixed temperature and this is start moving It was a small cell and we use it We were using only five milliwatts laser which was a surprise at that time because everyone was using already one What or put laser ruby lasers then and If we look at the mid plane with PAB we could see this this is for visualization Well, and this was the real image. We were using at that time and when the Temperature was increasing we see this we can see that the patterns were different So these ones was described as two roles in this direction And this was was described at two roles cross by two roles because at that time people were using just a Doppler velocimetry or Projection Sadography or something like that. Okay, so we end up knowing that it was something like this in this case There was like a toroidal flow Everything was going up in the middle and going down in the corners or so But nobody knew because the computers at the time were not powerful enough How were they streamline the 3d stream less or that flow? okay, so in this case It was the first time that we show that with this DSPI or the DIPH technique we can measure the in-plane velocities this and this comes from this image and Then this was the other one and this was the out-of-plane So you see this is not symmetrical the beam was the sensitivity vector The illumination was in this direction Vx and observation was MVY Okay, so the component was a mixture of this and this I don't see this This and this okay, and the asymmetry in this image This image is due to the out-of-plane velocity Okay, so that was the first time that we use this for measuring the three components in this box This was was so small 25 millimeters by 12 millimeters I have that you cannot do a stereoscopy at all you cannot do a stereoscopy at all Okay, so then what else we did at that time This image I have showed you the Fourier transform, but this is what is called a spatial filtering okay, I Can do optically a filtering of this image and get immediately Isolize for the horizontal component the vertical component. Okay, and then is what this is standard But then because the flow was steady We started saying well this looks like a two-dimensional flow is it two-dimensional or not well Yeah, then let's just calculate the 2d streamlines and then we see there are spares They are not circles if they don't close together it means there is an out-of-plane component because the streamlines has to be close so we have measurements accurate enough That when we do this derivative and integration path we could see they were not 2d even in the middle nobody was expecting this they expected that near the wall There was some transition, but they expected a big region in the middle that was to the measure and it was not so then we Well, let's go and do a three-day analysis. We took several images for several planes 15 or 17 place alone the Perpendicular composition and then we use the continuity equation for calculating the out-of-plane component because we didn't know about the Interferometry anything about it by then and then In this case we could have the three components 3c on the whole volume 3d And also we will have the limits that were velocity zero in the in the box Completely zero and then we can calculate the 3d streamlines And this is what you want This kind of thing the flow is going like this Let's go to the middle Megabit spiral and go backs from the outside if this distance is longer The outside is bigger if this distance is smaller The radius is more similar. So this is the only way to fill out This area with 3d streamlines that don't cross each other because they cannot cross and then we check that for the three-dimensional Area is was exactly the same pattern So at that time we could calculate where how the fluid was really moving and how the three-dry string like would like look like Look like okay, then computers came the people from Tarragona, Rovina will give work with them They did a simulation and they came to our lab to check on their simulation and they they proved this they But what else we can do with this kind of technique well You see this How do you think this is obtained just by placing a reference beam in a different point One, two, three, four, five, six reference beams The easy way to do this with six reference beam is to use a fiber optics multiplexer So we did this with a continuous laser because we tried with the pulse lasers, but When the pulse with is five nanoseconds the energy is so huge that the fibers destroy So I've been waiting for a long time for the hollow core fiber optics to be developed To see if we can use it with high speed laser, okay And you see that we always have the two images just always On the opposite direction going through here one one prime two and two prime But then we can say one two three four five and six and then we could recover the face From each of the six planes independently So we could do what I did plane by plane on the convective cell We could do it simultaneously. We could record this simultaneously and Recover it independently and we could do this in a flow that is steady Or it's not periodic. If it is periodic you can move it a long time So this is an example on the same cell with just two because two we can do this too with just Optics not the fiber optics, so they look different if they were not reconstructed independently we would see nothing with something Okay So now what else what are all other problems other problems we have is that When we use the usual first laser like this one, this is the one that goes up to 10,000 pulses per second Although usually for 10 1000 pulses per second The coherence of the laser is not good enough Okay on other lasers like the ones that go at 10 pulses per second or so they put something Which is called an etalon and they can increase this to centimeters or even meters, but I have I haven't seen this done on a high-speed laser So what we discovered was that the coherence length is about 10 millimeters and so because the beam is traveling In this direction on the fluid we never ever can recover the whole area at the same time If it was a solid because the light travels like this I could I could record the whole Area of the solid object because the phase the distance the light has traveled to the surface will be exactly the same But here is increasing so in this example the light was coming from here to there So we thought well what we can do well in the reference beam I can overlap three four five six whatever number of beams I want each with a different optical path and the optical path is changed But going through different amounts of glass because the refractive in the glass is different So we have this setup on the lab which is here and once you have it aligned it You can you can take it from a laser to another to another and put it wherever you want to So this is something that artificially enlarge the coherence length so Here is an example if we use only one of these paths which is our zero We only see these images. You can see the Fourier transfer here You see the images of aperture you see the face map and you see the intensity When you use the next one, which is a bit longer in panel, which is something inside Maybe farther on farther on the fluid so with this example if you put all of them together you get the full image of the of the The algorithm was in this case and you see here the Fourier transfer in the whole image Okay, so this is the only way to access the face in an object like this that is only two centimeters What we have been doing now in our flow, which is no biology guys increase it and until 80 centimeters and we Have been successful doing certain things Okay, so just keep going with the vascular flows This is the annular example in this case. I'm showing you some results While from two planes plane A and plane B. They are perpendicular to the main flow. The main flow is in this direction Okay, so this is the faces images Filters smooth they are nice and As I told you In each pixel I have information about the one velocity component This is the in plane components measure with PAB and The combination of these with this gives me the perpendicular this one And here the same Okay, this is not as solid as this because the horizontal component is contributing so on to this face map This is the little bit in convenience if I could make then Completely perpendicular with perspective you cannot you cannot do it. You cannot do it okay, and This is the simulation versus experimental Data this was done by Salvatore Cito the son of you know Well, he was at the Robira and Virginia University. He came to our lab to take the data from PAB from PAB He only compared really the PAB in a certain position He was the one that told me go the difference inlet and outlet means from the clinical point of view you have up there Ontario carotid artery left in route All these names that you probably know more than me about this and this is an image from the same orientation So this is PAB and this is a simulation. I think they are Colors here means the absolute value of the velocity No one company that's what they are very similar They use these for a challenge where there were several groups doing simulation then comparing the results He also showing that paper this Which is this plane a perpendicular one I am showing this which was taken in a different moment to Show you that they are quite similar. It's not a solid the same position One of the difficulties of this comparison is that you need to know well. What is the shape of the model? They have been measuring the same with MII I think they take this model to the clinic and take the data and then makes it We also try to do it by imaging several parallel planes, but then it was difficult to compose Okay, so these are some results on this example And this is the flexible one Okay The problem with these silicon ones is that they really don't make them patient specific I think they could but usually they make something like this like a round blow or so So this is example of the velocity vector map. This is the two components from PAB In this case we use two cameras. We look from the two sides because usually the data When you do PAB on one camera and this on the other camera are a little bit better There is a little bit of noise and then the combination. And one thing I want to show you in this example is Do you see this line here? You see I have lost everything here. Do you realize the time intervals? You see for PAB it was 2 milliseconds For interferometry it was 5 microseconds And here I am putting in 100 microseconds So what do you think this could be? Two images. The flow was constant in theory It was a pump with two valves pushing like this. It was not a piston pump. So what do you think this could mean? Any idea? These are the walls. What do you think this could mean? It is wall deformation It is wall deformation The wall, the vessel, the pump was having very very little changes in pressure And we repeat this experiment with the piston pump No fringes here But the other pump was just two valves that were pushing at different frequencies when you wanted to measure the Change the flow rate. So they were pumping. So they were introducing very very little changes on pressure So it means very very small changes on the deformation of the walls Because these were different points of this pressure valve. So it was repeatable So there is more information out here. Here you don't see anything because there are too many fringes Too many fringes and then you don't solve them Okay, and the last technique I'm going to be talking to you is holography We have already talked about image holography. This is going to be holography in general When you think of holography people are within 3D. I can get everything So in this case we are focusing on volume illumination and of course with intensity and phase Volume illumination. We could do it in this direction and it has been done, but From for digital recording is much easier if you do it in this direction Because you almost have to change nothing. Nothing. In fact, you see here. This is like a photograph You don't need to change anything and You see how the Lohola look like When you sense a laser and this can be done also with maybe just with a white light You don't have or with a lead. You don't have high Requirements on gohiness because everything goes along the same path. Okay, so what you do is you always see this kind of thing a dust A little dust you have on your lens will give you these fringes This is the 80 pattern rain fringes a particle you flow plane will give you these fringes Anything will give you these fringes. Just if you eliminate and look you will see that with or without lens. Okay so in this case what happened is that the face cannot be used as before on an interferometric analysis the face it was allows you to Recover the position of the particles so when we go to this ball illumination we are reconstructing particles and we are tracking particles so we're sort of using the correlation analysis to improve a little bit the Particle displacement, but what is more important is the particle position because what happened is that in this case in the direction of the optical axis the accuracy of The particle position is worse than on the transversal position because when Because the aperture of the image is not too huge a particle which is spherical or adult will transform in something like this Something like a cigar so we have a big area the intensity is not the same But we have a big area to look to place the particles so this is the main problem. Okay, so here you can see This is a whole lot of particles are worse until 150 microns or so. They are really big. This was a two-phase flow. Okay So if we reconstruct the image is what it will cause it as there is where they are more or less focused this particle look like this But three millimeters away You could still see something like a particle so the main problem with the software and this is what we are working now also is in Detecting where is this particle because then I can place there the Information on the velocity of my voice in this point not three millimeters away in one direction or three meters away in the other direction Okay, but because we are reconstructing particle and using intensity our ruler No, because We are reconstructing and after taking the the sensor we go back to the object space and do the reconstruction here Okay, so again, it's obvious space the rule is the object space but because the analysis is being done on particle position is not on Lambda lambda is no the ruler the ruler is the pixel position on the object space if the camera is 10 microns if I look in at the very small area it means that it's It's a pixel we corresponds to one micron in object space. So this is the ruler the Pixel size in an obvious space Okay, and what we have proposed recently is a way to avoid the the Virtual image when you do the reconstruction you don't see it here But this is the focus image which can be the real image and Focus I mean big big big rings very faint still here They seem to not be seen but when you calculate the displacement they contribute to the displacement I will show you so our proposal was why don't you use a Fourier transfer hologram To remove one of them and this is what we did three years ago. Okay in this case This is illumination and Here this is only for expanding the beam here between the lens and the sensor on the focal plane Where the illuminating beam is? focus We put the aperture and left pass only half of the Fourier plane Plus the dot because this is the reference being and you see the reference being is this origin So our Fourier transform will recover the information on the Fourier plane, which is this information here Okay, so this shows one of the aperture is the real and what is the virtual if we don't put anything both will be here overlapped Okay, so our ball is also to increase the amount of particles we can put in the flow because that will increase the spatial resolution So these are some examples with a glass plate covered with particles on one side. This is what it looked like Normal digital image holography. You see all the rings. They are perfect. And this is how they look like You see the only half of this in and they can be like this or like this And it means if they are like this they can be they will be in front of this of the hologram if they are like this It will be behind the hologram So this tell us some information also about the Z position and this is the position because position is the most important Problem. This is positions recovered from the usual digital hologram. This is the focus is image The real image and this is the virtual image. So if I go to the symmetric plane I also get the particles it is in focus, but it's in a different position So I have to select one or the other and you see many of them are Go to the middle because they know if they are looking at this or this one and this is what we get We are completely removed the other thing and then what happens? This is the position when this move if this point moves to the right This one also moved to the right, but if this point moves up this one moved down So the not only the position in Z, but the velocity in Z is a huge problem on this setup It's not here. This is what it shows here at this placement here On the normal ones there are problems here in our system. They are not Okay, so now I think this one is a nice one Again a flow Although this one is not being modeled by anyone is just checking the technique Can't you see something on the left? Something moving So these are the particles on the flow So what is the rest of it? The rest is that we have the bulk silicon that always has something Something in the bulk so everything that changes density will give you a pattern Okay, so how can we remove this? Well because that is static and the flow is not static we take 200 image Average then all and subtract the intensity from the hologram and then these all orange Change into these ones. Well, this is this has been also Rescale the intensity for better show so these are the number of particles we are looking at so we are trying to get Positions and velocities from this This has been done The analysis one would you go or so and here is what you have So the we place the the artery the carotid artery in such a way that the motion is on the plane X Y There is nothing on between or in the in the perpendicular only and X Y. Okay, so here Julia Lovera, which is the expert on this technique is plotting The position of the particles and the displacement in colors This is the velocity component the velocity modulus the velocity on the plane Okay in colors the colors mean the velocity. Well, as you know This is has to be up or safe flow in the middle will be brighter The velocity will be bigger and near the words you will bring the other or and because there is a verification here The velocity will be bigger here than here and here this is equal to this so more or less Here we have red and also green and whatever because we have the whole volume the whole depth Okay, and this is a perpendicular plane Two millimeters thickness all the data that are on this sheet of two million are plotted in here So what would you expect here regarding colors and dots? Will you expect the dots moving in this direction or not? Z is perpendicular will They shouldn't move the velocity is zero and then what we expect here Well, we do the spec here high velocity here, and then another circle with a different another circle with a better one Because it's what a post a postage flow is this it's like a parabola a parabola it. Okay So this is without aperture and this is with aperture. You see this area is a little bit better. There are less Strakes of these which means that the Z is Better the Z position is better recover, but we are still working on it and trying to prove it and we are using this just as a test Okay and then Well, I think maybe we're gonna stop here. I will not talk anything about the endoscopes You have questions. I think it's about time missing it Thank you very much. Thank you very much for this interesting and especially very enthusiastic talk It's nice to listen to you and especially with the examples Maybe a little bit of a practical question say that one of us wants to do a simulation and we say we want to validate it in Order to do this type of experiments What does it take is this you have a model and you measure it and an hour later You have the results or what's the practicality is how long does it take in order to do this type of images for me? To get the data with PIB You mean everything is set up. I take the images and then there are data Well, I'm first also to make this setup It's like to is everything is this something that out of the box You say like just give the object and we put it there and we measure and or is it a lot of work? Like this type of images to get like 3d. Okay. Okay. Let's stop three techniques PA is very easy Very easy. It's like you just you can do it even with your phone You take your phone you look at your area and you only have to change magnification Because there are no requirements on coherence or whatever. Okay, so this is easy then if you want to do Interferometry and you have a laser with a short coherence that can take two days to set up and in fact we have now it's set up for this mixing flow from the Rovira and Virgil University and To change it to be used with one of the ceremonies will take us weeks Even though now I have because we have to The path may be two meters and you have to match it within one millimeter We now have a lot of students it takes a long time And for this the setup easy to the same as PIV nothing no problem. Okay, then analysis So PIV is easy to set up and easy to analyze because you can do your own software and Malla you can buy it from a Company or you can even Find it for free on the internet that are so that's easy and running it very fast So that's the fastest the quickest and it can give you some information Analysis from the IPH one is done. It's always fast one the setup is done you take the images the same time it takes you to take the The PIV and the analysis is also on the same time. The only thing is that we are the only ones doing that Nobody wants to do it La vision long time ago the first time we discovered that we were in an European project and they told us They show some interest on commercializing it, but for whatever reason, I think they thought it was too difficult to set up so and then this one Easy to set up and for this one we are running the analysis on a cluster Not in a normal computer because it takes more Analyzing and we are still fighting to find out which is the base way to detect the position. There are different Things you can try so Julio Lovera is the expert in this one and takes Maybe one day one day and in which situations do you want to use which techniques? Okay, the way we have this setup You can only use it because of the way the whole around is in for for sizes of about the size of the sensor in Taspersal in depth wherever you want, but you have to take into account that you have 2000 by 2000 pixel you have four millions points plus Multiplied by the grade level you have did some amount of the function So you have something which is very deep very deep then you may have a lot of particles So you have to decrease your seeding so then the resolution can change and also another thing is that Even if you don't want to look at the whole depth If you illuminate in the volume and the whole depth is with particles You will see the know because there is no way to remove it So if you can do this Example on air and you can see only part of it you could do that You could use a small area and a huge depth because you will look only a certain area But I think this can be also modified to look at bigger areas Add in one more lens or so it could be done So you are asking about sizes okay for this one and the other one PIV anyone any size you want They have they have done it also on way on big when tunnels The only thing again is the same if you look up at the big area Your spatial resolution might not be good enough So you have to have a compromise in wind tunnels what they do is they put a lot of cameras Because you have the laser you can put a lot of cameras a lot of Laser if you want to and you can look at wherever you want in in muscular flows Normal you don't have such big flow. So I think you could look at the whole area you want and They are peach, you know with these four four beans four times I Enlarging the cojones lead you can look at three four centimeters. So and You also mentioned that you could see sometimes the wall deformation would have a vessel So is it in principle possible to measure and the flow pattern and the wall deformation at the same time or Can you extract the information from it? We didn't extract it because That was not a priority. So we have the images. They are very nice We think we know how to do it that we haven't done it, okay? the only the other thing I forgot to tell you is that We also saw the formation on the on the vena cava wall Because you remember I saw you the edge detection. Okay, so when we put the pulsar time flow The edge were moving a bit again we didn't analyze it because the student finished the psd She was not interested in that part and she didn't finish it But you could see in that direction. It means that with the vena cava filter with the data in a cover the deformation Perpendic in this direction was big There were big chains and pressures in the annulism and the deformation was not anything that should be there And in fact, we also put these pulsar type flow through the annulism But because we ordered it very very thin because someone from here told me I have an annulism Doesn't the form it was so thin that when I put the pulsar type flow the annulism collapse Because the change of pressure and I was afraid it will explode. So I didn't put that flow in there And the other questions Thanks for the talk it was really interesting. I was wondering So the examples you showed are done on Transparent models where visible light can penetrate. I was wondering whether these techniques could be used on models or even on People with other types of electromagnetic waves that could penetrate or is this something that could be applied to okay I was surprised yesterday when when your third talk They were showing you a lot of velocity fields. It was velocity fields. I Was going to ask the speaker if all of them were on real people because as far as I know PAB was being used was we use with Ultrasound in a technique that is called echo PAB The way the ultrasound works I suspect is Because I don't remember well. I haven't had time to review this paper. I read them a long time ago I think they were sort of integrating the velocity along the path because you usually put the ultrasound here And you have the detector here. Okay, but I think they claim they could do something similar to PAB But I was surprised if he was using it at a clinical level because I didn't think it was a technique at a clinical level But I think it can be this can be Another thing I have read also from a people I knew was working on PAB that they started a company for doing x-ray Holography Not just PAB but also holography But I haven't seen it being used But then they are I don't think the DIG page will be done. The echo PAB is when you inject contrast bubbles So you inject little gas bubbles and then you can track them with ultrasound if you do fast imaging Yeah, but the image the problem is only is like the resolution the use of contrast so in that sense It's not super accurate. Yes, but the bubbles occupy the whole vessel Well in the heart, no, it's quite small bubbles. So you can see them quite well. Well, no But no, but what I mean is that if you can put them only on the middle plane Then you could measure the velocity on the middle plane because in a vessel, which is Sylliptical only the middle plane is important. The other one is you can assume that but you cannot do that and I think with the ultrasound what they measure is the change in Frequency which means change in positions and fill it up. So then I think they have the the integration. I'm not sure But that is true. The contrast is not the same but the the speaker just did all was quite happy Portain any other questions? Thank you for your presentation and what about employing the laser Doppler anemometry or philosophy entry instead of the holography for a measuring Velocity the question is What about employing the laser the laser Doppler laser Doppler anemometry is a technique much older than PAB I think 10 years old or something like that because at some point they celebrate the 25 years or one and 35 years of the other technique the problem well I know Doppler velocity is used a lot on the clinic With ultrasounds I think and why they're using this because the Doppler velocity only need one pixel One sensor and one of the problem with ultrasound is that the sensor for ultrasounds are not as big They don't have as much pixel at the sensor for optical so you can do this Doppler with one But you get only formation on one point And this is I think this is really this that is used on the clinic a lot. I think Okay, any other questions Okay, thank you very much and then now it's coffee time. Thank you. Thank you