 Okay, good morning everybody. I think one of the exercises that we can do like every other course that I've been participating in, there is an effective half-life for the attendance. And it goes down, I mean, so as time goes on, you know, the attendance, particularly in the morning. Okay, so good morning everybody. Hopefully, I mean, it's not too early. What I want to do is this morning to give a little overview about the concepts and implementation of stereotactic radio surgery. It's not going to be possible to get everything about radio surgery in one hour, but at least what I want to do is to give you some ideas about the things that are specific to radio surgery. Now, how many of you are doing radio surgery? One, two, three, okay, good. So any questions that you have, you can ask them, okay? So let's get started, okay? Basically, I'll just cover a few things, basically some concepts about the targets and those distributions which are specific to radio surgery. But nowadays, as we will see, I mean, there is gradual overlap because some of the things that we are going to see in radio surgery are not very distinct from other 3D and IMRT, 3D conformal IMRT. So what I want to do is give a little bit of an history so you know where we started and where we are now. I'll probably talk about, very briefly, about commissioning and quality control issues, some issues specific about image fusion and target delineation that are important for radio surgery. Some of the different delivery methods and probably will go and illustrate it through the radio surgery treatment process. So where did we start? In 1951, a neurosurgeon in Sweden, Lars Lexel came up with this idea that instead of doing surgery, just cutting the patient open and taking care of lesions in the patient, perhaps a feasible idea would be to give a high dose of radiation and to ablate the tissue without having to open the patient up, basically, okay? So that was the concept. It was not radiation therapy. It was basically surgery. You want to kill the tissue, okay? So the concept, it's quite different from what we have been talking all along. You'll have all this in your handout so it's interesting. I mean, there is a very nice interesting article that you can download for free from Tim Solberg and others about this that gives a little bit of the history. But if we have to define what is radio surgery, it's basically both using a stereotactic system and we'll talk about that in a moment and high energy beams to irradiate that particular volume. So it requires two things, two components. Volume-based defined volume which has its index to a stereotactic system physically that we can point to any point in space in that volume and that we deliver the radiation to that volume using the same stereotactic system. So that's the two things that are a little bit different than what we have been doing. And the other characteristics is that both of them are designed to produce as sharp a falloff of the radiation as we can. So what is the stereotactic localization basically? It's, we have a stereotactic head frame that it's rigidly attached to the patient's skull. So mechanically, this space around the patient's skull is just one unit. And any point within that space that we define can be referenced to a specific coordinate in that space. That's basic concept and very simple. Now any point or any structures should be able to define them and in that space and use that to both define them and treat them. So what is basically a stereotactic localization? If you take that space and here we have a diagram of a reference frame, which basically it's attached mechanically to the patient's skull through that base. And we can cross take cross sections through that. And the thing that is specific about that, if you see this set of rods and there are different configurations of basically localizing rods, but if you look at the cut through this top of the head, you see that there is very little of a head in this space and there is this sets of points which are the cross sections of this rod. Now, as you can see and just pay attention to the ones that are highlighted in yellow, as we move through the different planes, that position or that point, it's going to move in relation to this other two that are basically your reference things. You can see that this point, the distance from this point to those keeps increasing as we move down. That's why this has a shape of a letter N if you want or so on. So, once we have that any point in that space or any cross section or any anatomic landmark can be defined by an X, Y, Z coordinate in that coordinate system. And the X and Y are very simple. I mean, this is your X and this is your Y. You have to be careful because depending on what system you have, the origin of that X and Y coordinate could be different. Sometimes you go to a CT scanner and it defines zero, zero at the top left, the top right, at the bottom. So, this is something that to be very careful. So, on each cross section we have X and Y and it remains to be defined what's the Z coordinate of that point. And the Z coordinate is given by in this case the angle of that bar at the diagonal. And from that point we can define the Z. And it's a very simple algorithm to come out with those coordinates. So, what is this treatment strategy? We position that point or the volume that we defined at the point of convergence of our set of beams and we want the intersection of all beams to be at the specific point that we target. Where does this thing started? It didn't start in radiation therapy. This is a historical picture of some of the different stereotactic frames that neurosurgeons used to have where they attach the patient in the operating room and they have a mechanism that is different mechanism, very complicated, that they could go and drill into the patient's head to reach a particular point. That's the origin of these stereotactic frames. Historically, I mean, Dr. Lexel introduced the concept of ablating the lesion by a single procedure like in surgery. And initially they use a 200 kV X-ray tube and you can see the mount for the X-ray tube mounted into one of those stereotactic frames. Basically pointing to a specific point. Later on he developed a gamma knife. It was the first gamma knife. It was consisted of almost 180 cobalt sources in a hemispherical arrangement. You can see the old machine here, which covered essentially a good portion of the space around the patient. So the newer gamma knife, basically, it's just the refinement of that system. The patient gets connected mechanically to this head support. This sometimes they use what's called a helmet with collimators because the cobalt sources not always can be made very small. So basically it's like a secondary collimator which is around the patient's head. Some of these can be replaced with different diameters, but basically you have a fixed geometry. It's not a gantry rotation or anything, which allows it to have a very good mechanical stability and precision. I am by training a machinist. I was a machinist before I went to physics. So I know that you can get with mechanisms, mechanic, very rigid mechanism. You can get precision of a few microns if you want. So a lot of times, I mean, we hear the gamma knife is precise to less than a millimeter, a quarter of a millimeter, a tenth of a millimeter. Yes, all that mechanism perhaps can be made to be precise to a quarter of a millimeter, but let's talk about that a little bit later in terms of what's the implication for the precision of the treatment. So the collimators can be of different diameters. As I say, once the patient is attached to this, this rolls into the head of the machine. They open the shutters, the patient gets irradiated and that's done. So you have an arrangement of beams. All used to cross at one single point, which gives you very nice spherical distribution of those. The initial uses of radio surgery were not for tumors, were for arteriovenous malformations. If any of you has had any experience with plumbing, you know, pipes and so on, well, that's what happens in the brain. We have the arteries, we have the veins. The arteries become smaller and smaller. They deliver the oxygen to the tissue and then the veins start collecting the blood without oxygen and drains. So normally there is a lot of impedance or pressure gradient between the artery and the vein. Some malformations are such basically that there is no microscopic vessels and the blood just goes through. I mean, there is no resistance. What happens, the tissue doesn't get the nutrition and the patient suffers the consequences. Typically arteriovenous malformations are pretty dangerous and once they happen, they can be deadly. So there was not very good alternative than to surgery and to go and cut off those shunts if you want. And with radio surgery basically they say, well, we know that if you give a high enough dose of radiation, the tissue will become hardened. And basically that's the idea. You harden those microscopic vessels and you create basically a scar. The only problem was that how do you tell where you have to irradiate? Well, typically they used to have radiographs with two orthogonal views in which you can see they put some dye so you can see the vessels and here you have an area where there is leakage and you can see it here. So from those two views you have to decide where you want the radiation to be. Later on with digital structure and geography you could do that with a digital system so it was much faster and you can avoid high contrast, the inclusion of contrast. So in order to have a reference frame you could have basically two views and in this case the images were registered to each other and you could tell the coordinates from those roads and the markers. The problem is that with two dimensional views you don't have three dimensional information. We saw that couple of lectures at the beginning of the week. So just as an example, if you have an object which is like this, like a half moon if you want, you project it into two directions, lateral and AP or whatever and what do you get? You get these areas. But from these areas alone you cannot reconstruct this structure in space. So you have only partial information. So how do we translate the location of whatever we decide that we are going to treat? That was the early stages. Into the stereotactic frame. We have these boxes with fiducials. I don't know if you can see them but there are fiducials that will show up on a radiograph and you can then take coordinates and you will come up with a coordinate which is more or less the center of that blurb of material and project it to the center. But you didn't really have three dimensional information. So just say let me choose this as a center. This is the volume that I want to treat. It's this size by this size by this size. So this is one of the, a little bit newer gamma knife units as I mentioned that. So what happens if you had a volume which was not just a sphere? You wanted to cover something that does this shape. Well with the gamma knife, the way to do it, you couldn't treat except little spots. So you started covering that volume with individual shots. You take one shot, covers that area, take another shot, move the patient to another coordinate, take another shot and keep on building your dose distribution like this. Now we all know that if you start doing like this with very high gradient dose distributions, like look like a big pyramid if you want, tall pyramid, probably more like an obelisk in some places, you cannot cover something homogeneously because you have very big gradients and you are trying to patch them up. And there is no technical way of avoiding that. For these purposes where you wanted to ablate tissue, probably didn't matter much. As long as you had a big sharp gradient on the outside, what happened inside didn't matter much. If you exceeded the dose per screen by three or four times, not three or four percent, three or four times, it was still, you ablated the tissue and it was okay. Clinically it was okay. So later on, people started developing linear accelerators to reproduce this gamma knife. First of all, the gamma knife was limited and it could only treat patients which where you could do this in one shot into the machine. You couldn't do multiple treatments because you couldn't reproduce the position on the frame. I mean you had to take the frame for the patient to go home or maybe just you could do a second one the next morning and keep the patient overnight. But that was about it. So people started to adapt the idea of the gamma knife to linear accelerators. Why? The main reason was that there were not that many patients that needed this procedure. And if you had a linear accelerator, you couldn't dedicate a linear accelerator just for that but you would do something that you modify the linear accelerator just for one treatment maybe once a month. For the X knife was one of the developments, early developments of commercial development in which you could do a number of things. You could do use collimators which were also circular collimators which they tried to mimic what you had on the gamma knife. And you could do a variety of things more than with the gamma knife. And you could do also lesions outside of the cranium and not only outside of the cranium. Initially you could only treat this part because that's the part that could go into the machine. But now you could expand to do a little lower lesions also. Later on we will see that we did it, et cetera. So these are the clinical results. Before radio surgery you have all this blood that is leaking here. And six months later, I mean you achieved basically control of that thing. So obviously the thing. So it is historically the first, I believe it's the first, linux-based radio surgery system and it was developed in Buenos Aires. I don't know if you know that. You do, okay. So it was a very strange contraption. If you can see the patient was sitting here and instead of on a regular couch, the patient could rotate like this, okay. And then you could rotate it on the base. You can see it's like on the couch rotation and the accelerator could rotate around and you can see the amount of the special collimator onto the gantry, onto the head of the machine. The delivery techniques with stereotactic radio surgery nowadays are done in a multitude of ways and we will just cover a little bit of a detail with that. First of all arcs with circular collimators which were supposed to mimic what the gamma knife was doing. We can do conformal beams. We do arcs with conformally shaped beams that they shape as we rotate around the target and later on even IMRT. So with the linux with cones, now we can do more than one isocenter of course. I mean that was feasible with the gamma knife as well. But typically the first treatments this was a technique that was developed at the Harvard Joint Center in Boston. You would treat basically with arcs that went, we were used to treat with arcs around the body but now you could put the patient onto the couch, rotate the couch and have arcs that go in different slices if you want. And basically once you do that you can make all those convergences onto a smaller volume. This is a diagram of the mechanics and that was the technique that was initially developed. You had an arc which was essentially on a cross plane and then three more, three more, four more arcs, three more arcs, sorry, okay, out of plane. So how did you create those things? Usually the collimators, you have to remember this is in the 1980s. The machine didn't come with the MLCs or anything. I mean that was developed much later. So in order to get a sharp enumbra, we did two things. First of all, we created the tertiary collimator, X and Y Jaws that they're secondary collimator. You have the primary and you create a tertiary collimator where you have a small aperture made of lead, usually precisely machined with a mount that mounts to the head of the machine. So this is a tertiary collimator. And this is one of the first commercial ones, so you can see here. This is the beam that is hitting the tertiary collimator. The collimator in some cases was made of two pieces and the final collimation was the lower one. And this part was mounted into the head of the machine from what I remember this is a Siemens machine. I recognize the collimator mount, that's the way I know it. And you have to define your field with the X and Y Jaws to just hit the center of that collimator. Remember that picture. So this is defined by your field. Okay. And then people manufacture a variety of collimators with different diameters and here you have some scans. You can see how sharp of a penumbra. Look at the scale here. This is one centimeter from here to here. Okay, so it's a penumbra of about a couple of millimeters. The collimator assembly therefore had to be mounted very precisely because any change in angulation will ruin the penumbra. And also will give you an isocenter. Now the isocenter is produced or defined by the projection of those cones. So how did you align that? Usually you had some kind of target which you mount on an XYZ mount that you can move precisely with 10th of millimeter precision. And first of all, you use your light field. So in this case, you can see there is a little, I don't know if you can see, there is a ball. You can see the shadow here. This is your light field through the cone. And this is radiating this little ball. This is probably a piece of film. And you can see it here. This is the spherical ball. And if you can move that precisely, you can make adjustments or use that as a reference in the space of the room. Okay, now you are rotating the collimator around it to test how accurate your rotation is in relation to the room coordinate. So, and you, I'm sorry, once you have that aligned, you can probably align your lasers to hit that little sphere, which will tell you this is the isocenter of the machine in the room. So the QA for the stereotactic system, many of you have heard about the Lutz Quality Assurance Tool or Lutz Test, yeah? Anybody didn't hear about that? You didn't hear about it? Okay, so you need to read about that, okay? It's basically a test that says, well, if I'm radiating this object in space, I'm radiating in these directions and the object is fixed in space. I don't move that. And I rotate the machine. I can look at where is that object in relation to something that is related to the machine? In this case would be the projection from the collimator, light or x-rays? Typically, it's used the x-ray beam. And the position of that target in space in relation to the image that you created through the collimator, you can look at the coordinates of that in different directions. It's just a geometric exercise. And you can say, well, this is off by three-tenths of a millimeter in the gun target direction up and down in the room and so on. The trick on these things, nowadays there is sophisticated software that you can do it all. It will calculate it for you, but it can be done manually as long as you keep track. What are the directions? And you have to concentrate on having just one system of coordinates because you'll have an image that will correspond to some point in space and you need to refer everything either to the room coordinates or to the machine coordinates, okay? But this is basically, so I'm not going to go into detail. You can probably read in that many articles and this is an example of software that does it for you automatically or semi-automatically as long as you know that you took your images in the right sequence that the software expects you to take, okay? Because if you took the first one with this gantry and the second one with this and the other one like that, it's not the same. The images will be in different order so the algorithm will not know which direction you were coming. It's just looking at the piece of film like this. And you can tell, you could probably measure physically on a piece of film where is the center of that collimator field and where is the shadow of that target, okay? And here you can see the variations and you have the automatic edge detection in this case and so on, so it makes it easier. But we used to do that by hand. I mean, it's not so hard. So what is the idea of doing this multiple beams, multiple arcs? They're just going to give a small illustration. In this case, this is where the micro-MLC but the concept is the same. If you treat this target with one beam, you have this distribution. If you start adding beams like this is a lateral, you get a better coverage, okay? We spoke about this in my first talk the other day, yeah? You start using multiple directions and if you do two parallel poles, I mean, you get even better coverage and if you start adding more directions, you'll get a very nice coverage. And if you now, on top of that, do this in different planes, now you get a real good coverage. Look how sharp this covers your target, okay? And very nice follow-ups. So this is basically the whole idea of stereotactic research. Multiple planes, multiple arcs, okay? And here is a comparison of the dose distribution. Both of them cover nicely but with this one, the 50, well, that's a 30%, that's the blue one. The blue is between 40 and 60% you can see it's much bigger for fixed, nine fixed fields versus nine arcs, okay? So when do we use circular arcs? Well, if you have a very small target, that's ideal. You have a very small, very well-defined target. With one shot, you will cover it and you are done. Typical examples, small metastases in the brain while they are still small. Trigeminal neuralgia is one of the areas where people have been trying to push the limit. And the reason this is pushing the limit is because, yes, you can put a four millimeter cone into the head of the machine. But your field is going to be, and I will show the distributions for very small fields. The uncertainty of where your field is in relation to how well can you target could make the difference between hit or miss, okay? Because we are talking now about sub-millimeter precision. So spherical targets, you could do up to nine arcs or six or seven arcs with the collimator of the correct diameter. But what happens if you have an elliptical target? Because not everything in the body is a sphere. So if you have an elliptical target, you can do a couple of tricks or techniques. First of all, you can just eliminate some of those fields or those arcs that go perpendicular to the major axis. If you have an ellipse, something like this, and I want to do arcs, I probably shouldn't bring arcs that are coming from here because I'm eroding in and out. It's much better to come with arcs that I'm pointing at this. But if we want to still make arcs and we want several of them, it's not all or nothing. So we will just select a few arcs that are more directed to the long axis, along the long axis of the target. You can see it here. We eliminated some of the fields. The other thing you can do, if you have a target like this, you could do the arcs that are in this plane and eliminate those. So that's one strategy. The other strategy is to use different diameters. The problem with that is that now between arc and arc, you have to go back to the room, change the cones for a different size, and that takes time and it's patient. It's not particularly comfortable, but that you can do also. So for circular collimators, these are examples of distributions. You can get distributions which are even like this with circular collimators. What other techniques can we use? Depending, so we looked at the circular cones and arcs which look the best, I mean the best penumbra. It's very precise. There is a very low integral dose to the brain. And the other technique that we started doing with targets that were a little bit larger, and you can see the range here between three and six centimeter, that was a rule of thumb. You could take the cone projection and some systems allow you to plan that and close the jaws within the field. So if this was like a three centimeter cone, you may have put the jaws here to give you a centimeter and a half in this direction and the rest is blocked by the jaws. So that's another way to do it. Or you can use a MLC or a micro MLC. This came out much later than when we started doing these techniques. This usually will result when you don't need super duper conformation to the target. If there is something that it's not very critical next to it or so on. This is an example of a complete arc. That's the dose distribution in space from the complete arc. And we see this target and we can select, for instance, this is a critical structure. We can select the position of the jaws. I'm sorry, the graphic is not very good here. But you can leave the X jaws open and the Y jaws, you close them to avoid this part. In some planning systems, you can just go and you look from the B inside view and you just separate one from the other with your jaw. And this is an example. In a standard arc, we get this distribution with a shape or conformal arc. We call it conformal, but it was really using the jaws. You can get this flatter dose distribution around the target. So what happens when you go to really larger, lesions, probably cones are not very useful unless you go to multiple isocenters. Well, if you are going into multiple isocenters, you are back to the situation with the gamma knife, okay? Covering this in pieces. So it's much better to use static fields which are conformal but not coplanar. So you still have the culture rotation and you can treat that. Micro-MLC probably it's the best option. It's best if you can reach your target from about two pi sterillion. So what happens when we collimate with the micro-MLC? This is obviously, you can see by the pixelation, it's a very large, magnified target. It's probably about only three, four centimeters by, maybe three, four or five centimeters by three. And you start putting an MLC around it. You know that you have two techniques for MLC use. One of them is to bring the leaves halfway into the field, other ones just to keep them outside. And there is advantages and disadvantages to do either of them. But let's say that we keep them a little bit out, just touching the sphere. And we use a different collimator size. This would be a centimeter by centimeter. So your resolution is such that you may have quite a bit of normal tissue irradiated here. So as you reduce the leaf width, you are better able to close that. So the use of micro-MLCs, initially the micro-MLCs were not built into the machine were add-ons, like a tertiary collimator to the machine. This is an example of one that was, I don't remember the brand on this one. I think it's, it's was Brainlabs, yeah? Yeah. And you can see that it's add-on. This was pretty heavy piece of equipment. Had to be mounted to the machine just to mount it. You needed the cart with the mechanism to lift it and so on. So these are the parameters, nothing special about that. And the key thing is you have to mount it on your tray holder. Now, you all know that the tray holder is not the very most rigid attachment. So just putting it to, as if it was a tray, it's definitely not enough. You have to have some way of pinning it mechanically with essentially sub millimeter or a tenth of a millimeter precision so you can lock it and tighten it in place. That's very important. So the, usually they had mechanisms to do that, that guarantee that. So how do we do is now we don't have the cons. We have the MLC shaping our fields. You can do exactly the same thing with the Lutz, that's the phantom, okay? The Lutz phantom. So the same test just now you have small fields and now what is the procedure typically for a radio surgery? You have the MRI images typically because CT is usually not good enough for contrast resolution in the head, inside the brain. You prepare the patient, you have a ring placed on the patient. This is a, you can see it's an invasive ring so it goes with pins into the skull. Not particularly comfortable but it's not the end of the world. You have to be careful that this is really rigidly attached because I have seen cases where you have these pins that go into the skull and you tighten them up to a certain pressure but I have seen situations where they were put in the wrong place. First of all, you have to make sure you avoid blood vessels. You don't get some hemorrhage from the hole but also that you come in a way that is mechanically stable. I have seen cases where one of these was put almost like at an angle to the skull and just moved during the process because you do it in the morning, you treat the patient in the afternoon. Patient is waiting while you do the plans and so on QA and by the time you bring it to the machine it was in a different position but you couldn't tell because it was on the skin and they had some swelling of the skin so you couldn't see it. So this is a different type of localizer which is not invasive and this was developed at the Harvard Joint Center also in Boston where you replace the pins now with a system that has an occipital mold to keep the head in relation to that. I mean, this is to take the weight and to make it a little more stable and I'll show you there is one more part. There is also this bite block which secures the position because the lower jaw is movable but the upper jaw is part of the skull. So if you have something that you bite into it and it's well made, you basically have a number of points and you apply pressure. Now you know who this guy is? This is Wendell Lutz, the one that invented that Lutz test, okay? He was just testing this and demonstrating how to mount this, okay? All colleague of mine, bright, bright man. So with a relocatable system, now you have the patient being supported here that is made custom for the patient. You want to have as little possibility of movement. Now this is less rigid and precise than something which is invasive. There is no question whatsoever about that. But we are talking now about fractionated radiosurgery. It's therapy, it's not ablation anymore. So that imprecision, I mean, it's something that over a period of five or 10 or sometimes more treatments will be. So how do we know that that frame is positioned on the patient's head the same way every day, okay? We put it once and now the patient goes back and it bites again into the mold and so on. Well, there is devices that you can make that are geometrically rigid. You mount them onto the base that the patient is supported by and you make those measurements which is basically the distance between this sphere and the patient's cranium. And if you go in all directions, you can get a pretty good error detection of the order of a millimeter to two millimeters. But you have to do it several times. You do it before you put it, before you start your process, throughout the process, before you do the treatment. So this is patient immobilization. That's a different type of immobilization mask, but there is a number of them now in the market. This is the key. I mean that the setup errors can be up to several millimeters. How this is made, I mean you have a mask that it's made custom for the occipital part of the head and then you make strips that secure it and they're rigid as much as you can. And then on top of that, you put another mask so the patient is mostly stable. Now, when you do angiography, sometimes I mean you don't, you have the angiography done without the mask, obviously, that sometimes it's done days before. And you can either do a CT scan with the localizer frame or with the angiography similar to the other one and you need to relate those coordinates. Now what happens? Now you bring the patient to the treatment area, you did the plan, everything looks great. Now you bring it to the, in this case, it's with a system that has what's called the localizer target frame and you can print these templates that go onto that frame. Now this space now is the stereotactic space and you can tell that there is some patient identifier, obviously you don't want to take one from a different patient so it's specific to each patient and you can have the marks that correspond to the isocenter line so you can align this with the lasers initially, that's a first approximation and you can see for each arc or for each fixed shape beam we'll project it onto those drawings on the box. Now what happens if you do that and now you take a port film or a radiograph of some sort and you find that, well, I need to move but I need to move your six tenths of a millimeter in one direction and two millimeters in the other direction. Now you have to have something that you can control the patient position before you treat. Now, there is a number of ways to do that. Most of the holders that hold the patient in place have veneers that you can just make very well calibrated motions or there is another system that has a set of two cameras that point to the patient or take images of the patient with reflecting spheres. If you took those spheres in your CT's set and you know where they are and this gives you the images in the room coordinates now it can tell where that frame, this frame is now in space and since it is attached to the patient you can make small corrections. This is a set of, it's not very clear from your distance probably but this is a set of x-rays in two orthogonal or two orthogonal directions and you can compare them to your reference images and it will tell you, well, just move this direction so much and this direction so much. Now, because this can give you the coordinates of that frame in space you can move the couch and you verify that with that infrared camera. So the question is how do you translate images from one system to another and we spoke about that briefly the other day. What we call image fusion in reality most of the time is image registration not really fusion because you have the information superimposed from one system to another and if it was important to know that in IMRT and 3D conformal, here is critical because the precision of your treatment will depend completely on how well you identify the target in one imaging system and you move it to the other coordinate system for the CT. If you have an error of two millimeters forget the fact that the gamma knife said you have a 10th of a millimeter precision in the machine you don't know the target to better than two millimeters. That's something to be remembered so this fusion algorithms are very important and you have to obviously verify them with much better precision or a position that is suitable to what you are doing. So you may have structures which are very complex. Here is your structure on the CT that's from the MR and you have margins as you do in conformal, et cetera. Your margins probably will be much smaller because there is much less room for something to move inside the patient. So you are really relying on the precision of your registration. Here you have a case where you have critical structures which are just nearby what you want to treat. I'm going to skip go through this. If you look there is a report 54 of the APM which is number of years old now but it gives you the different factors what affects the precision that you can expect from the system. So if the importance of this is this one the CT slice thickness and nowadays you can do CTs with multi detector CT and so on with slice thicknesses of a fraction of a millimeter or a millimeter. When we started typically you didn't do any more than maybe two or three millimeters and that they will also be a problem because your image sets were huge and this is the most important thing. If you see all the contributions to the uncertainties the stereotactic frame, the alignment of your isocenter and other things tissue motion and so on the one that is critical is the CT resolution. If you have a CT which is three millimeters thick that's where your resolution is being limited by. You cannot tell anything better than probably about half the slice of the CT. So planning with conformal arcs. I mean just you want to. So I'll try to skip some of these. You can always look at that. There is different ways of planning. I think what I want to really focus more than anything else is issues of QA. The comparison between dynamic or conformal arcs versus multipliesocenters is an illustration of what you had in terms of those distribution. Yes you can do cover very well but with multipliesocenters you are giving much much higher those inside. And if you are talking about therapy that's not trivial. The QA of the isocenter I mean this you need to verify that periodically on the machine that things are correct. I want to skip that to some of these. The pymdosymmetry of small fields and I'm not going to go at all into that. There is a whole series of publications at the difficulty of getting accurate those in very small fields. And the factors that you need to measure in order to put your data into treatment planning system are here. But the critical one is this. This is for gamma knife but it's the same thing for small cones. If you have a small cone about the smallest here is about four millimeters. That's your dose distribution and look at the scale. This is half a centimeter. So basically you don't have an area where you have a flat beam to measure to begin with. So let's say you put your detector and your detector is four millimeters wide and that's not a big detector. That's not a far better. You are covering basically your detector is going from here to here. So part of your detector is looking at the different dose fluence. Correct? So what is the result of that? First of all, you have effects of lack of buildup. So as you look at your TMR curves or percent-depth doses they are going to be significantly different just going from one cone to another. And more than that the output of your machine is this. It goes severely, severely down. Look at the scale here. This is about one here and this is about 15% lower. And they have been comparisons where very sophisticated institutions measure output and for small cones, look at this difference. The same person or the same group measure this relative output factor, 0.3. They re-measure it 0.7. So if you did that, what happens to your patient? If you made an error of this type of magnitude on your output factor, what happened to your patient? You either over those or under those by a factor of two. Not two percent, two, huh? Okay, so this is a very critical thing. Anybody that is even thinking about doing radio surgery better understand what they are doing in terms of the absolute dosimetry, okay? I cannot go into that because Eugenia will just keep me out of here. So then you have issues with the verification of your dose calculation because we are talking about small fields. The calculation algorithms are not perfect, okay? And you need to do patient specific measurements. Measurements, I'm not going to go into this one, but you have to do those volume, those evaluation surface, those summary size, those volume histograms. I'm going to skip that one. You need to define quality indices for your treatment, but you need to do this. If you are going to do radio surgery, this is an absolute must before you treat first patient. Treat a phantom, which is anatomically designed, and have somebody review that totally independent from you, both in terms of the dose and the dose distribution. This is a report from the RPC. Now it's called the IROC at MD Anderson, and you can get that done wherever you are in the world. They will agree to send it to you. You have to return it, of course, with all the measurements done, but you treat this as if it was a real patient from beginning to end. You submit it. You have a TLDs and film now inside the phantom, and they will tell you, this is the dose that you reported. This is the dose that we measured. Okay, so we say, well, yes, you did it good, you did it bad, and where is the error? This is a list of, this is from another paper that you can download, typical errors in radio surgery, major accidents in radio surgery and the causes. And one of them, remember I told you about the collimator and the field size heating that collimator? Okay, a major accident happened where somebody was setting the machine, put the collimator, somebody says, let me set the field size. The therapist sets the field size, or the physicist sets the field size. One or the other, one tells the other, well, what's the field size? It says 40 by 40. The other one says 40 by 40. One meant 40 millimeters by 40 millimeters. The other says 40 by 40 is a 40 centimeters by 40 centimeters. So all this collimator, it's just a waste of your time. The beam was open, completely delivered. You know, thousands of rats to the patient outside of the target. Okay, a total disaster, okay? So, QA communication is extremely important. Task group 54 has a little bit of a flow chart where it shows that, you know, what are the things that can go wrong along this sequence? One of them is to set the collimator. In this case was six centimeters by five centimeters. Whatever it is, somebody has to double check it before you do anything. That's probably the worst thing that you can do. All the rest is, you know, okay. So, thank you for all that and for your attention. I'm sorry about being the first one in the morning. Okay.