 So, okay, let's start with this second talk, which is devoted to organ motion management. And as I told you before, I will also use some concepts we have already seen in my first presentation here. When we speak of organ motion in radiation oncology, it's quite clear to all, I think that we speak especially of respiratory motion. But we have to be aware that we have different sources of motion in radiation therapy. Spiritual motion is of course the most important. It's a pseudo-regular motion. It's not completely regular, as all we know. It's predictable in a very short interval if needed. And I'm going to show you later why we think that it's necessary to predict the respiratory motion in some applications. But we also have skeletal muscular motion, which is an irregular, completely regular motion that can be controlled, asking the patient to cooperate and using a positioning device that you have seen in the previous presentation. We also have cardiac motion, which is generally not explicitly accounted for in radiation therapy, okay? We have gastrointestinal motion, which is unpredictable, totally unpredictable, but can be partly limited. And we have motion of the genital urinary system, in particular bladder filling, which is important in prostate treatment, for example. And especially if we use hyperfractionation techniques that take a very long time. For example, I have a direct personal experience with hyperfractionation in prostate that was given in fractions of 40 to 45, even 60 minutes sometimes. And in that time, of course, we have a problem of bladder filling that results in a generally a rotation of the prostate of several degrees. It can reach up to seven, eight, or even nine, or 10 degrees. It can be partly limited with the cooperation and the equation with the patient that has, once again, you have heard in the previous presentation. But for my part now, I have this choice. I will devote my time today with you on just respiratory motion, okay? And in particular, I will do it by analyzing the principal lines that we see in this important document of the double APM, which is management of respiratory motion and radiation oncology, which is not particularly new now. It was published in the year 2006, but still is a very important report which is called the report of the TG-764 for respiratory motion and radiation therapy. So we start by having a look at what anatomical sites we are dealing with when we talk of respiratory motion. Of course, we expect respiratory motion in lung. And for shallow breathing, which means superficial breathing, not forced, okay? We have descriptions in the literature of motion that can reach up to five centimeters for the lung. For the esophagus as well, we have important displacements. We have displacements of the liver up to four centimeters with normal respiration. We have the movement of the pancreas, which is also very important. Pancreas is thought to be an organ that is quite fixed, very deep in the anatomy, but still it's very subject to motion due to respiration. Three to 3.5 centimeters is a normal range of motion in normal respiration for the pancreas, which is really a problem because it's quite close to very critical structures. So if we have an approach of high doses in radiotherapy for the pancreas, which is required because pancreas does not really respond well to low doses. It's really a problem of toxicity for the nearby tissues. The breast moves with respiration. Even the prostate moves with respiration, even if this is generally not accounted for in any way. But I think, for example, or some techniques which fortunately are not common that treated the prostate in the prone position. If you put the patient in the prone position, you have a more severe problem of displacement due to respiration than in the supine position. So you have to be aware of that, of course. And for kidney as well, you can reach several millimeters. This is a table taken from double APM test group 76 that shows you the displacement report from the literature for lung lesions. And it is divided in three columns, which are superior, inferior, motion, interior, posterior, and lateral motion. As we expect, the most important motion of lung lesions is in the superior, inferior direction like this, because it is due to the motion that the diaphragm, okay, which in respiration makes the lesion go down, and in respiration makes the lesion go up. This is the most important. And as you can see, the various studies that are reported here give a displacement of nine to three, nine millimeters, sorry, to three centimeters. This up to five centimeters, three and a half centimeters, and so on. In the anterior, posterior direction, it's less important, but still it's a big range of motion, can reach up to two to 2.5 centimeters, and it is very observed for lesions that are in the upper part of the lung, okay? If you think of a person which is breathing, normally you have an anterior, posterior displacement like this, in this part, but when you go down, you are close to the diaphragm, the motion that you expect is in the superior, inferior direction like this. And this is important if you describe the motion with an external monitor, for example, which always goes practically up and down in the anterior, posterior direction, okay? Please be aware, do not think that any external monitor is describing exactly the direction of motion of what is really inside. This is a similar table for abdominal regions, the pancreas, liver, and kidney, and you can see here that motion of the pancreas is reported up to 3.5 centimeters, which is an important range of motion, of course. The sources of information that we can use to measure or to account for motion in imaging are the one reported here. We have radiography, of course, a double exposure or a cinema modality. We have fluoroscopy with or without fiducial markers that can give us some information on the pattern of motion of a lesion, if it's quite visible. We have ultrasound, but the most important tool that we have is once again CT, and in particular for DCT if available, that is able to describe in a complete, an entire cycle of respiration, giving you the information of different volumes pertaining to the different instants within the respiratory cycle. We will see a movie in a while describing this. We also have MR, and for DMR, but these are very less common techniques, very difficult to be used, and we have seen before, we have PET and 4D PET, which is already quite common in various centers now. Of course, patient training is very important if we have to deal with motion. This, for example, is a graph which is taken from a study coming from 2006, which shows you the variability of the respiratory patterns in free breathing with no patient instructions and cooperation compared to an analogous cohort of patients for which instructions and audio visual feedback was used. And it's quite clear how the variability is much more restricted if you use things like this. Other strategies can be the administration of oxygen, for example, to allow patients that breathe with difficulty to be more cooperative and to maintain a predictable pattern of respiration. The double APM-TG76 report that we have seen gives you several levels of motion management, starting from motion encompassing techniques which are the most easy to do, going up to respiratory gating in increasing complexity in this way, and then with breath hold techniques, forced shallow breathing techniques, and the probably most refined from the technological point of view, which is tracking or respiration synchronized techniques. I will try to show you some considerations on at least three of these levels which are motion encompassing, gating, and tracking in increasing the level of complexity. For motion encompassing, motion encompassing means that we estimate the position of a lesion, a target, like this one, this red dot one, for example, in say, expiration. Then we estimate the position of the same lesion in inspiration, for example, on the two extreme ends of respiratory cycle, and then we build what is called an ITV, an internal target volume defined in this document here. Which is built putting together the clinical target volume with the internal margin, and can be represented, for example, with this dotted line here. And on this volume, you build those distribution that encompasses the whole trajectory of the lesion, which is taken during the motion due to respiration. Of course, the main limitation of this study, it is quite easy to do, even if you do not have a real 4D CT, okay, with the cooperation of the patients, a normal respiration and a CT taken in full inspiration and full expiration during normal breathing, you can actually estimate the range of motion like this. The main limitation of this is that, of course, it takes a lot of volume. You have to irradiate a lot of a healthy tissue to be sure that your dose distribution, which is static, covers something that is actually moving within it, okay? So you're not sparing that much organs of risk and the surround destruction by using this technique. But it's easy, it's readily available in practical all centers, and allows you to miss the target in small fractions of cases. This is an example of motion encompassing. This is taken from a real treatment that was performed some years ago and here you see the situation of the lesion in this sagittal plane, which I suggest you to see rather than the other ones, okay? This is the position of the lesion in full expiration. This is the dose distribution that is built around the lesion to cover it, and this is the situation in full inspiration, okay? So in the inspiration, you see that the lesion moved here, in expiration moves here, the dose distribution is fixed, and the lesion moves within the dose distribution, okay? This is the concept of motion encompassing technique. Well, motion encompassing techniques, of course, still requires margins for uncertainty in position, okay? Do not confuse this one with the concept of passing from CTV to PTV for setup errors. Setup errors are to be accounted for separately for this motion, okay? And this is the concept of gating. Well, the main limitation of motion encompassing was that we had the whole trajectory and we had to move the lesion within a large dose distribution, okay? So why not using, for example, a fixed position like this and switching the beam off when the lesion is moving like this, okay? This is the concept of respiratory gating in treatment delivery. We deliver the treatment just in the phase that we have taken in our city. We have planned the treatment in this phase, and then we have a system at the accelerator which is able to shut the beam off when the position is not the position that we use for treatment planning, okay? This is a respiratory gating. Of course, the main advantage of this is that we can spare volume in this case compared to the other one. The limitation is that we get very long treatment times. It depends, of course, on the fraction of the respiratory cycle that we want to use to do this. If we want to use 25%, for example, the respiratory cycle, we have treatment times that require four times the beam on time compared to the motion encompassing technique. If we use, because we want to be more precise, just one in 10 phases, we, of course, have 10% of beam on time, which is useful, so we have times which can be 10 times longer for treatment delivery, which, of course, is a problem from the clinical point of view. But still, this is the concept of respiratory gating. Sorry. Respiratory gating is made, performed by means of systems like this one. You need an external marker, which can be a block like this one, this thing from an external camera, okay? And it's important to understand what you are looking at is something that is not really the tumor. You are looking to an object that normally goes up and down in the anterior posterior direction like this, while the tumor very often moves like this in the superior inferior, so you have to be aware of this and especially have to be aware of problems like the one reported here. Here you have on the left a case where the motion of the external marker that you use on the chest or abdomen of the patient is described by the green line. The motion of the tumor is described by the blue line, okay? In the left panel, they are very superimposed to each other, so if you use the marker motion to give the beam on time, you are quite sure that you are irradiating the tumor when the tumor is within the window, good for treatment, which is enclosed by these two lines here. But if the correlation between the external motion and the motion of the tumor is bad or is not constant in time, you can get a situation like this one where you get the beam on time corresponding to the marker motion within the window, but this situation does not correspond to the tumor motion within the window, so you are irradiating the tumor or you are missing the tumor because it's not really in the window that it was planned for the treatment. So this is another important limitation of gating together with the limitation due to long treatment time. Well, and this is the concept of tracking. Tracking is not very user-friendly and as we use today, we have just a very few systems that can really do a tracking, a respiratory tracking for radiation therapy, but the concept is to spare the volume together with time by redirecting in real time the beam while the tumor moves like this. You build a dose distribution around the target and then you redirect, you move the beams to move the dose distribution while the target moves. This is done, for example, by this machine here which is based on an industrial robot on top of which is mounted a linear accelerator and by using the whole degrees of freedom of the robot, the system is able to follow in real time the motion of the tumor. This is one of the very few systems available today for tumor tracking, but there are studies for using multi-lift collimator, for example, for tracking which probably will become available in shortly, in probably a few months or few years. What is the problem of tracking? The problem of respiratory tracking is, probably you have already seen in my previous slide, that if you build a dose distribution around your target here, and if you have an organ at risk here which might be the spinal cord, for example, you make a dose distribution by planning in one single phase and then you switch on the system and the actual dose distribution is not what you planned but involves the organs at risk that are around the lesion. So the solution to this is that you need a real 4D planning. It's not like in motion encompassing or ingating where you just need a simple CT for simulation. Here you need a real 4D CT because you need a description of all the respiratory phases and you have to build a dose distribution which takes into account all the situations in the various respiratory phases that you have. And for example, there has been the evidence in the literature in the past of a real deformation of the dose distribution due to this effect here. So it's really a thing to be made with much attention. Okay, problem in motion control include also imaging, of course. We have to be sure that we are using good imaging, proper imaging to describe the situation that we are going to find at treatment. In particular, we need a temporal coherence between imaging for treatment planning and treatment delivery treatment administration. Imaging shall always describe the treatment condition. This is very important. And this is especially important. We have already seen in the previous talk something if we use quantitative imaging to describe the tumor or to describe the type that we want to irradiate. And in particular, if you use PET CT to describe a biological target volume, a BTV. This is what you get in CT if you do not control respiration. You see here that we have motion artifacts. I'm sure all of you are aware of this problem here. And this is the correct coronal slice corresponding to it. And this is the image of a sphere taken at CT in static condition and on top of a table that simulates the normal respiration. And of course, what you get is an object that does not have the volume and does not have the shape of the real object that you had to scan. And this is another important part. You can solve this problem in particular if you have a tracking system with 4D CT. 4D CT is a modality in which you have a CT scanner and an external reference like the blocks we have seen for gating before the accelerator that allows you to make in correspondence each single slice with the instant within the respiratory cycle in which that slice was taken. And this allows you to sort the slices to build N 3D volumes corresponding to the single respiratory cycle. This is what you need absolutely if you use tracking. But once again, if you just use gating or a motion encompassing technique, even a normal CT simulator or CT is enough to get good information. 4D CT can use a prospective or a retrospective acquisition. Very briefly, the prospective acquisition allows you to acquire the volume just in the respiratory cycle that is of interest to you. And this results in dose sparing, of course, but you have limited information on the respiratory cycle. Or you can use retrospective sorting in which you have a redundant acquisition of the volume continuously and sorting which is made just after the acquisition was taken. You have, of course, higher dose here but full information on the respiratory cycle and what you need if you use 4D planning for tracking, for example. Okay, this is just a concept I had already told you. Before. Yes, please. That is a strategy that is good, in my opinion. If you want to use MIP, MIP allows you to get a description of the whole range of motion, right? We do that in some cases. In some other cases, we use a real 4D CT and then we estimate the envelope of positions when we build the CTV. One important consideration to do is that you do not normally need a high number of phases to get a good result. For example, if you compare what happens with 11 phases, which is a standard, which is a standard, a gold standard, I would say, for the CT to six phases, you have an average dosimetric error, which is just 0.3%. If you compare 11 to two phases, you get 1.5%. If you compare 11 phases to an average phase, which is made averaging the two phases taken here, for example, you have 2% error. You make a big error just if you compare the 11 phases with just one instant in the respiratory cycle. So the message is do not aim for a high number of phases if you use 4D CT. Just be aware that resolving the respiratory cycle within different times, just two times is enough to get a good situation for the dosimetric point of view. For magnetic resonance, I would not spend much time on 4D MR because it's really not a modality which is already used for treatment planning today. Still, you can do breath hold here. You have some cinema mode or very short acquisitions like echoplano imaging that allows you to do very fast imaging and get the different instance of the respiratory cycle. Real 4D techniques are not so common and they can use external surrogate signals, for example, or internal surrogate signal like the position of the diaphragm and so on. But once again, it's really still in the research phase rather than a real clinical application. So if you allow me, I will skip this part on MR just to go to PET CT, which is much more important in the definition of the biological target volume to irradiate. This is really very much used today. That is a real quantitative imaging modality as we all know. We do not have just FDG, which is the common modality. I would say it's 99% of the use of PET CT today, but we also have applications in radiation therapy with different traces like fluoromisonidazole here, which is a marker of hypoxia together with copper ATSM and so on. All these modalities have in common a problem, which is the very low signal to noise ratio and the fact that motion can be really a problem for their use. I have already shown you this slide in the previous talk once again. It's a comparison between the SUV, maximum SUV that you get in this lesion here. If you do not take motion into account, here we have an SUV, which is below two, which is quite low, okay? Could be even mis-taken by for a benign lesion if you do not see the real situation here. But if you control motion, you have SUV values which are much higher. Once again, this is a very extreme case. Fortunately, not all the lung lesions behave like this, but it can be an example of what is really happening in reality and how motion control techniques can help you getting the real situation. Motion is of course a problem of quantification here because it has a huge impact on quantification. And the methods that we have today for control of motion impact are scanners like this, which are equipped with a 4D pad system. This is a variant system which is called RPM, which is the same system that in our center we use in one of the linear accelerators. So we have a couple of systems, one in PET-CT, the other one at the accelerator, and allows us to get for the information like this. This is another very extreme case which in my opinion is quite interesting. On top, we see a lesion which is this one, okay, taken with no motion control, which is in full acquisition during the whole breathing cycle. And here we see on the left the maximum inspiration and on the right the maximum exploration situations. You can see that here we do not see the lesion anymore because it's displaced inferiorly in inspiration by the motion of the diaphragm and the motion is still there in expiration. But you can also see two things that are very important. First, the boundary of the lesion is much more defined here than here. And the lesion is even bigger here than here. But on the other hand, we have a problem here and here which is the low signal to noise ratio. I'm sure you can see that the image quality, if you look at this region here, for example, in the mediasional region here, is much less defined than here. This is because with gating, we are losing a lot of information. If we use, for example, five phases and we get just maximum exploration like this, we are discarding 80% of our counts, okay. So the signal to noise ratio that we get is much worse here than here. And you must be aware of this. So one problem is how many phases do we use in 4D path to get a good information? Let's go back to this one. If we want to recover the full quantification for information, we must use nine to 10 phases or 11 like here, okay. But if we do like this, we must be aware that we have a problem of signal to noise ratio which might be very, very low. So maybe the optimum number of phases is around here. I will tell you what is the choice that we have made in our center. We, in our center, we have made this choice. We, if we need the 4D city, we just need, use five phases up to now. Why five? Because if we start, this is time. This is percentage of respiratory cycle from zero to 100, okay. Generally the systems that allows you to do a 4D path use 100% to describe full inspiration, okay. Then the respiratory cycle goes down like this at 50% of the time, yes, 50% time, you get full expiration, and then it goes up again and you get 100% which is once again full inspiration. So this point is the same point here, okay. You start from the beginning from here. The maximum expiration point is exactly in the middle of this graph. Well, if you use an uneven number of phases which might be four or six, you have a division of your graph that splits maximum inspiration into. If you use an odd number of phases like three or five or seven, you get a situation in which your maximum expiration is correctly described by wise one single phase like this. So the choice was between three, five, and seven, okay. Between three, five, and seven. If you look at this graph, if we use three, the signal that we get is quite low once again. If you use five, this is a very extreme case once again, but in the vast majority of cases, if we use five, we get an underestimation of the SUV which is no lower than 10% compared to the maximum, okay. So the choice was if we use four D-pad CT, we just use five phases. Of course there can be more refined strategies to use for the information and to recover the whole signal to noise ratio. For example, we could use one of the techniques of the formable registration that we have seen in the first talk. With the formable registration, we could do something like this. We have a four D CT data set here. So this is for example, maximum inspiration and this is maximum expiration say. We register each single phase to the three dimensional phase with no motion control and we get n maps of the formable registration. Then we take the whole path without motion control and we apply the inverse transformation to get the single phases in path. This is called the virtual four D-pad strategy and allows you to recover both the real shape and the SNR, the signal to noise ratio and the activity of the region. But once again these strategies are not like in the formable registration commonly implemented in commercial systems. It's still in the research stage. So you just have to be aware that it is probably coming in the next years but not yet available. For all techniques of organ imaging with organ motion, we have to rely on tools for quality control. One of the tools of quality control that we can use is the one that you see here. This is a motion phantom which has two important characteristics. First, it can describe both the motion of the surrogate signal which is the block here, the external signal and the motion of the real tumor inside the lung like this. And second, it can be driven with a real, for example, log file taken from a system that describes the respiratory motion of the patient and in this way it's able to describe a real motion and not just a mathematical model like a sinusoidal or something like that. These two characteristics are very important because they allow you to describe exactly what you get in the real patient. And of course, tools like these are not the only tool that you have for quality control of techniques that manage motion. You can also make an analysis of log files of the system, make consistency checks, for example, see if the volume is preserved in inspiration, in expiration to have a real description what is happening in your system and of course you cannot make anything without the expert judgment of a physician that says you if you are taking the wrong or the right direction. I would like to finish this part with an example of a decision tree or whether you have to use, it can be for the PET CT in our case because this is an example of a decision tree we use in our center, but you can as well apply it to for DCT without PET. To make the decision if we need for the technique or not, we first of all see if the anatomical region is affected by respiratory motion and the patient can tolerate the procedure. For the first part, the region affected by the respiratory motion we normally estimate five millimeters of displacement as a minimum. If the patient can tolerate the procedure and needs an explicit technique for the respiratory motion like this one, we estimate the lesion with caution. If the patient cannot tolerate the procedure of course we discard the idea. If the motion is greater than five millimeter, we make the decision to use for DCT or for the PET CT with five phases for the reason I have already told you. If the motion is smaller than five millimeter, we go to a standard planning imaging. And once we have the for the PET CT, we see if we have a close proximity between the target and the organ set risk. If there is a close proximity, we choose the phase which is the one that allows us to spare the organ set risk. If we have no proximity to the organ set risk, we use generally maximum expiration because it is the phase in which the patient spends more time during the breathing cycle and which is more reproducible compared to the other. And then the phase which is chosen is used to build the PTV and then the PTV with the setup error sensor. So in conclusion, tumor motion or organ motion must always be considered and accounted for, but it's not always necessary to do an explicit management of respiratory motion. Sometimes it's really too complicated, too long for the patient, too long for your clinical activity to be performed. Explicit methods for motion management might prolong treatment time and introduce significant uncertainties. For example, we have seen the uncertainty that we introduce if we do not take into account properly the full 4D information in tracking. We need a temporal coherence which is always necessary between imaging for planning and treatment administration. And we also need to put in place a quality assurance program that allows us to validate our technique and see if we are making any important mistake. And finally, it's always necessary to make a balance between accuracy and clinical applicability of the study which can be a very refined technique that allows you to do very nice things but maybe it's too long for your clinical work. So in my opinion, it's always necessary to do a proper balance between these two aspects. Okay, that's all and thank you very much for your attention once again. Thank you.