 I see that the lecture was very good one and that's also my opinion about that lecture because discussion is very hot. So that means that at least it touches in some way your brains. I hope you will keep concentrating until the end of my talk because I can imagine that you are at least a little bit tired because listening is also an effort, not only speaking but also listening is. Well, the title is commissioning of treatment planning systems. That's really important part of our work. Why? Because we treat our patients using the data we get from treatment planning systems. And that even can be very dangerous. Why? Because many young people treat a treatment planning system like as another game, computer game. And it is a game, but it's very serious game because you can really kill a patient if you don't use properly your treatment planning system. And we have such examples of killing patients by using wrongly a treatment planning system. And the accident, Panama accident, that's exactly the case showing us that system can be used in the wrong way by very ingenious physicists. So commissioning is a very important task and unluckily it's never ending story. Because every year you have to come back to commissioning of your treatment planning system. I will touch that problem later on because it's very serious problem. But again, before you start, it was very nice introduction to my talk. First of all is to understand the system you use. Understand means what you may and may not expect from the system. Professor Hartman told you that for example well two different algorithms. The first one in Pennsylvania and second one is collapse convolution algorithm. What is the main difference between these two algorithms? Where you can see the differences between these algorithms? Well considering this scattering is an issue but what in practice you will see the differences in lung. Exactly in lung. Where the transport of electrons is important. In pencil beam algorithm the transport of electrons is not treated properly. Of course it's treated much better than before. So as again remember that never our life is perfect. We go step by step I hope we go to the better solution. So collapse cone is better than pencil beam but might be the new algorithm which is implemented by variant into eclipse is better than collapse cone. Convolution method. But anyway you should understand what you may and you may not expect from your system. That's the first problem. The second one is problem of interpolation and extrapolation which is very often missed by people. You enter some set of data. Let's say you enter the depth doses for a range of beam between three and forty square fields. Three centimeter, forty centimeter square fields. And if you want to use your system for radio surgery where you very often use the beams of the size of two centimeters. You should ask yourself how the system will cope with that problem. Oh thank you very much for that comment. It's exactly the same for IMRT. If you use very small segments for IMRT. You should ask yourself how the system treats this data. Because extrapolation can be fully wrong. Well, you know that. If the person's dose is like this and you make the extrapolation. You know that there is a mistake. Let's say here is output factor. And this is size of the beam. We know that it's go like this. So that value differs very much from the real value. So well whenever you use the treatment planning system and you try to or start to use it out of the data you enter to the system. Be very careful. Bore's problem. You have air gaps. Somebody asked me about how to treat the problem. Because always we know that from our practice that we have some gap between Bore's and the patient. And that's real problem. Of course Bore's is just to create electrons. So if the gap is not very big one, doesn't matter. But anyway you should check that. Or you should realize that that's the problem. Okay. Do not expect good results without precisely measure input data. And that's dosimetry. If you well enter the wrong data you will get the wrong calculations. And well from my experience it's clear that people very often don't look carefully at the measure data. You should do that. Analyze your calculated data which is done in my department very carefully. Before the data are transferred from the dosimetry group to the planning group. Because my well center my radiotherapy is really very big one. We treat every year about 7000 new patients with 8 accelerators. So well the group of physicists also big one. So there is a dosimetry group. There is a planning group so we transfer the data. But we talk about the data. We analyze that. So for example of course the most important is how to how output factors were measured. Because that's the most important data for the treatment planning system. The second one is depth doses. Okay for example well don't forget that effective point of measurement for cylindrical chambers plates. A different point as the point of measurement for output factor. Sometimes people forget that. Because according to new protocol it has been changed. In the previous protocol 277 which is much better protocol in my opinion for physicists. That was exactly the same point. So you always have to move your chamber after setting the chamber at the place. Profiles. You should consider the finance size of the chamber. Because if you use let's say cylindrical chamber, farmer chamber for measurements of profiles. The penumbra is much wider than it really is. And of course your data cannot describe the reality that it should describe. So all these problems should be considered before you start to commission your treatment planning system. For example coming back to the output factors. When you prepare the table of output factors. You should graphically present the data to look whether there are smooth data. Along y-axis and x-axis. I strongly recommend you to make it. Immediately you will see that you made at least small mistakes, small errors in measurements. Of course you can smooth the data but do it also very carefully. Electron density conversion function. Very important. Because if you commission your system you always compare almost always. You compare your data measured in water phantom. So in homogeneous absorbent. So you just rely on calculations performed in homogeneous data. But you cannot have the good results if you don't have the right conversion. Function from electron density from Hansfield units to electron density. I know that not everybody has the specialized phantom for that. But in my country also not everybody has it. But we just, you know, we take the phantom from another center, use it for one to day. We measure the electron density curve and we enter the data to the system. That's really important. Okay, now that was introduction to my talk. What accuracy is needed in contemporary radiotherapy? You have TCP and NTCP. And you may see there is really a strong dependence between the outcome and the dose. So what we expect from radiotherapy is let's say 3.5% uncertainty of those delivery treated as one standard deviation. But I tell you that right now in a very good center we are at the level of 5 to 6%. So we still are far away from that. That's not because of the treatment planning system. Only partly due to the treatment planning system uncertainty of calculations. Mostly because of all these issues related to movement of the patient, setup of the patient, changing of the body of the patient and so on. But anyway, treatment planning system partly take parts in that uncertainty process. Here is definition of normalized dose gradient dimension to you so you may easier understand what it is. I don't want to focus on that. Well, commissioning of a TPS. Several reports were published concerning commissioning of TPS. APM, Dutch, Astro, IAEA. We have also published such recommendation. I don't want to criticize all these recommendations. But they are very complicated, very long, and it's really impossible to make all these tests they propose in these protocols. So in my talk I tell you what is the most important commissioning of the treatment planning system. Of course that is my personal view, but that's my experience I think can be also well accepted. Examples of these, they are accessible via internet so there is no problem to launch them and to read them. Just to be informed how the people propose to check the treatment planning system. Okay, sorry for leaving that slide. That also is important before you start commissioning of the treatment planning system, but not only for treatment planning system but also for any other activities. If you think about setup correction, you also before you start you should say what you accept and what you don't accept. There was a question during a break. What happened if two days I have two millimeter, three millimeter and the third day I have eight millimeter? What to do then? To calculate the mean value after three fractions or no. In our course we have the action levels. If it's larger than action level we have the special procedure for that. Before you start with treatment planning system you should also define what you accept and what you don't accept. And you should define what you do if you don't accept such a data. Okay, so there is range of acceptability and not acceptability maybe you may define three different ranges. But that depends on you how to deal with this. What we should remember, what you should remember that you always rely on comparison of measurements and calculations. But whenever you measure, you measure the actual, let's say output factors, PDD output factors. Let's focus at output factors, okay? But when you measure output factors you measure the actual output factors. Which not need to be exactly the same as the output factors you measured before when you measure the data to enter into treatment planning system. So you should correct your data for actual value of your output factors of your accelerator. And here is how to make it, it's very simple. It's just to multiply your result by the ratio of, by the ratio of output factor. You enter into treatment planning system and actual value, okay? Because, well, radical is saying that if your output factor should be one centigrade per one monitor unit. And today you have two centigrade per one monitor unit, of course you have big difference. But it's not because there is a difference between treatment planning system and, and your measurements. But because the accelerator doesn't work properly, okay? So you should correct it. Tolerances can be expressed in terms of dose, but that should be applied in the low gradient region. When you compare doses and in terms of distances in high gradient regions, sometimes the concept of gamma is used. Well, gamma concept is a little bit complicated and I think it's better for commissioning to use these two definitions. In low gradient dose and high gradient distances between point in which the dose is exactly the same. So that's the proposal of IAEA, how to treat it. So in inner space you use dose comparison. At penumbra you use the distance comparison and outside of penumbra you use again the dose comparison is used, okay? So, but you have to decide before you start where and which data used to compare the data of measurements and data of calculations. Here are the more detailed description of that. It's difficult to read it, but it doesn't matter. You can look into the document which is accessible in the Internet so you can read it carefully. That's the really difficult question, complex question, overall accessibility of the treatment planning system. I just read, a useful way to compare calculation and measurement is to analyze the deviation statistically. So not if you get one wrong result, you just not try to remove your system from your department. Also a given tolerance may be assigned to individual point value comparisons. The decision of overall acceptability is not based on strict adherence to the tolerance at each point. Rather decision are based on confidence limit or other similar criteria. For example, a few points may fail to meet the tolerance of 2%, but this may be acceptable if 95% of points fall within 2%. Okay, my comment to that is as follows. You cannot change your treatment planning system. If you buy Eclipse, you have to use Eclipse. If you buy Monaco, you have to use Monaco. If you buy XIO or some of the both XIO for you, you have to use XIO, okay? Your aim is to check where and why you can expect this on that results. And you will understand that in that situation, you cannot expect better result, that result you got. And for example, you should say to your doctors, I don't want to use that system for that particular situation because I don't believe we got the right result. For example, somebody plays the ball and say, well, but I want to have that ball 4 cm away from the skin surface. You say, well, but please consider that it's very likely that my treatment planning system doesn't calculate the dose distribution for that situation properly. Oh, if somebody wants to use your treatment planning system for radio surgery with beam of 1 cm, you say no. I don't agree with that because you know I cannot rely on my treatment planning system. If somebody proposed you to calculate the dose distribution for total body radiation and you haven't checked whether your treatment planning system is calculating the dose distribution for 2 or 3 m distance properly, don't use your treatment planning system. Before you do that, check your treatment planning system. That's your responsibility, okay? Please remember about that. Well, for more detailed discussion of tolerance, please refer to IAA document or directly to very good Venselhar work, which is also how possible to get from the internet, however, it might be green journal is not fully open journal. Now it's again a very important question concerning the treatment planning system and that's why I told you that that's never ending story. Because the question is when to make and which test should be made for your treatment planning system. Certainly when you enter the new data to your treatment planning system, you have to check everything. For example, in my department right now, we decided to measure again the full set of data for one of the machines. Why for one of the machines? Because we have several, let's say, identical machines. We try to keep five of our CD to 2,300 CD accelerators exactly with the same data, the same output factors, depth doses, profiles. The differences we accept are less than, in terms of PDB, less than 0.5% in terms of profile, less than 2%. And output factor has been due at, let's say, a very tiny difference, okay? So we decided to measure again because we know that right now we have better detectors and we can measure more precisely our data. But before we use clinically this data, we enter the data and we fully commission the system, which is very time-consuming procedure. There is one medical physicist who is well obliged right now to make this commissioning of our treatment planning system. So whenever you have the new data, you have to perform, let's say, almost all tests. If it's new system, all tests has to be performed. But the problem is that if you get, for example, new software version of your treatment planning system, what to do then? That's really a very tough question. You cannot rely on your treatment planning system. What to do? First of all, please don't accept the new version of software very often. Accept the new software if there is really big change in the software, which makes your system much better. If it doesn't make your system much better, there's no reason to. For example, the colors are nicer. Okay, that's the case. Treatment planning system is not lady. It shouldn't be very nice. It should be very good. So please be careful with that. But anyway, if you have changed the software, you should analyze which part of the software has been changed and how it can influence on your calculations. And what is very important, never leave any, let's say, information which tells you that it's something wrong in the system. Because we have such attitude to our life that if we think that something goes wrong, we think now it's just by chance. Next time will be better. That's our philosophy. In that case, you should treat seriously all signs that something is wrong. New accelerator, that's another case. If you really set that new accelerator exactly as accelerator you had before, which I strongly recommend, it's much better to buy exactly the same accelerator, however it's not always possible. For example, Siemens disappeared from the market, so if somebody has Siemens accelerators, he will have to buy another accelerator, another type of accelerator. So then, well, try to keep the data exactly the same. And then you have to make but only part of tests. You should look at the data you enter and compare the enter data with the data you measured, but that's, I think, enough to check the treatment planning system. Well, what to start with, I have already told you about that, but again, the most important is to be sure that there is no errors in input data. So you enter the data into the treatment planning system, you calculate the dose distribution in the water phantom, and you have to compare measurement with calculations. And that is a must. That is a must. So if you enter, let's say, 10 depth doses, you should calculate 10 depth doses, and you should compare them. And the first step is to compare them just visually, by eye. Okay, I don't know how to say. Visualized inspection. Thank you very much. Well, because that's a very useful tool, just to look and to see whether there are big differences between what you calculated and what you measured. Next, you come to the numerical comparison of the data. And sometimes if something went to the system, it's very good just again to compare visually, compare the calculated and measured data. In most systems, we enter output factors, depth doses, and profiles. Sometimes you also enter the attenuation factor for your trace, if you use individual blocks. And that's also important to check that carefully. As I mentioned, visual verification of the data is obligatory. This is flattened filter fleet. Profiles, which makes us really a problem when we compare it, because you have to compare carefully this not very flat curve. So it's difficult to say where is low gradient and high gradient after beam. So that makes the problem. Okay, depth doses versus field sizes you should also compare. And as I mentioned, output factors. And that's the most important part of your work. After third part, you can go to another one. Just to make some tests to calculate spatial situations and compare results of your calculations with your measurements. So the first step is always to input the water fountain. If you measure some data in the plastic water fountain, you should scan your plastic water fountain. Never use the fountain which was entered manually. Just scan it via CT, enter into the system, and then calculate those distributions in that fountain if you want to compare such a data. Okay, and also very important as functional tests. These are qualitative tests that ensure that the right interpretation of the treatment plan parameters are get. For example, it's really crucial point convention of angles. So if you rotate right, that your angles increase. If you rotate left, your angles decreased and so on. Okay, and I've seen such situations that people made a mistake. Luckily, they check it before they use it clinically, but that's really very dangerous because you wanted to irradiate your patient on your left side, but the system shows you that you should irradiate that patient on the right side. But we have very serious problems related to the positioning of the patient on the CT because it can be head first, feet first, first left, right and so on. That should be checked very carefully from the very beginning to the end. So build a very simple fountain with the one, let's say, left side marked and head marked. Place your fountain on the CT, scan the fountain, rotate the fountain, scan the fountain, rotate the fountain, scan the fountain, send the data to the treatment planning system and compare whether interpretation of the data are correct. And that's really important. That might be one of the most important points. Many mistakes are made if we rotate the table, the couch. That also should be checked. Well, checking the treatment planning system is not only the checking of the treatment planning system, but the connection between your treatment planning system and your verify system because you send the data to a verification system and you should be sure that everything is right. So, well, it might be a very general message to you. As a physicist, you don't work only at the computer in the treatment planning group. You should have very strong contact with patients and with exerators. That's your responsibility to go there and check whether everything is right. Naming of collimator jobs, naming of orientation of wedges, for example, variant machine has eight different wedges. It's really very easy to make a mistake and to get lost using these wedges. So before you start to use, just prepare the plans of your treatment planning system, send to the record and verify system, go to the accelerator, check whether everything is right. That's clear. Oh, check of input and output image data, input CT and digitally reconstructed radiographs. I cannot tell you how to make it, especially concerning digitally reconstructed radiographs. But in my career, we found the error on Centra Master Plan system where there was error related to reconstruction of DRR and thanks to very, very strict work of my colleagues from my department, they noticed that there is something is wrong. They just made portal control and they noticed that there is a difference of, let's say, seven, eight millimeters. And according to our protocols, we tried to explain what went wrong. So we took that patient to the simulator, we repeat the procedure and we noticed that everything is right. So we thought, okay, it was just by chance. So patient again went to the accelerator, we got next portal control and again we had about seven to eight millimeter difference. Then we again went to simulator and again everything was right and then it came to us that something is wrong. And we start looking for a mistake and when we start to analyze the images reconstructed by the system, we took the phantom, anthropomorphic phantom, we prepared the plan, we make pictures, we were sure that everything is right because with rigid phantom it's quite easy to prepare the right geometry and then we understood that something is wrong. And we found that DRRs depends on the CT information. If CT information, well, the step between cross sections was, I don't remember, it was one millimeter, two millimeter, everything was right, but it was five millimeter. The system made a mistake and that mistake was corrected in the system. But that, well, please, the major message to you is, treat seriously the mistakes or the uncertainties you observe and try to explain them, not that everything is right, but try to find that not everything is right. When you then not find any reason to say that everything is right, then it's not right, then you are happy, okay, but that's anyway, that's really a problem. Qualitative test, check of input and output image data, okay, comparison of calculated measured dose. Some of these tests have already been performed and non-standard conditions. So you can measure in different situation, different geometrical situation and having the analyzer, so I mean the water phantom with automatic movement of the chamber, you can measure different, you can see more like different situation and you can measure dose distribution and in a moment I will show to you a proposal of some of such tests. For example, block geometry, you may measure the field size with block, in the beam and you measure the doses at certain points. Again, please remember that your measurements should be absolute one, not relative one because, well, you can imagine that for example, if you measure your profile, how much time I have? Sorry, I have no time, yes? Sorry, okay, let's imagine that if you have measured profile and profile is not absolute, usually it's not absolute value, okay, because you may measure also for accelerator in absolute values, then you may like this, calculate it, okay, it should be like this, but if you normalize it, you see there is no difference, but in fact there is. So always it should be the link between normalization and what you use for your comparison. Another proposal to make that test, just you can measure it with your phantom. Another one, okay, and we just performed this test where we put into operation our treatment and exist. Wedge beams also please take very carefully, inhomogeneities, that's the big problem. The first one is analysis of the algorithm, what is possible, what is not possible, and you may measure in the anthropomorphic phantom or in the slab phantom, you can simulate some situations, okay, but you never can simulate all situations you have in your real work. So you may also compare your calculations with the result published in well, recommendations and papers, but then use the data. If the quality index of your accelerator is not very different from the quality index, it's mentioned that for that accelerator the data were measured, okay. If the quality index is very similar one, you should expect very similar results for your accelerator, for your comparison. And stability of the TPS, I recommend to repeat the same procedure for the same anthropomorphic phantoms regularly. Let's say once a year, we repeat that procedure once a year. We have one plan. Once a year we calculate it again. We scan the phantom, send the data. We calculate dose distribution and we compare that dose distribution with the dose distribution and we calculate it once a year before, okay, to be sure that nothing happened to the system. It usually nothing happened, but it's, well, you know, it might happen. So please be careful with this. And very useful is so-called end-to-end tests. So you get the phantom, you examine the phantom on the CT, you counter the phantom, you prepare the plan, you calculate the dose distribution and then you irradiate the phantom with, I don't know, with films, with TLBs, with chamber placed somewhere if you can place the chamber and compare measurements with calculations. Of course, it's very time-consuming and, well, hard work to do, but it's better to have something to do that, do nothing. My guess, from time to time, it's nice to have nothing to do that. Anyway, okay? Okay, that's exactly the same idea I mentioned before. And, well, you may ask what to do in case of non-conventional machines. I cannot give you good advice. You just should follow the recommendation I gave to just normal machine, common machine, and you should yourself, well, prepare your test of your machines. And you should follow the proposal given by manufacturer in case of these machines. It's very complicated. To be honest, I don't have cyber knife in my department. I don't have tomo therapy, so for me it's difficult to talk about that. And summary of my talk, commissioning of TPs is a must, is a very labor-extensive and boring procedure. However, there are people that are very, you know, interesting in checking everything. Robustness depends not only on the system but also on the user, so the limitation of the system should be known. And, again, it's not recommended to change the version of the software because we are insisting to change the software because you pay some money for that. But it's not a good idea to do it because, you know, if system works, it works. So it's like with a good car you can rely on. If it goes, then don't change the car. Yeah? And please be especially careful when you start with using a new system because, well, especially if before you used another system because your experience is based on the previous system and you think and you expect that in that system everything will be the same. But it's never the same. Never the same. Very useful is to give students to use the system. Usually they find mistakes in the system because they are doing very strange things using the system and then everything crashes. Okay, thank you very much.