 Well, good morning to everybody. I well, I promised you to To give you reference to that book. It's that book It's as I mentioned, it's rather old book. It was published the the last edition was published in 1983 But I think still it is really a very good book from just physics Point of view how medical physics should look like So if you want to have the physics background for your work That book is really Is really nice is profitable for for you today when I looked for that boy I even found it in PDF in internet. So maybe it's not fair But anyway, that book is there Okay, it was very nice overview of Imaging systems which can be used for radiotherapy what Professor Hartman made but before I Will go to my lecture you ask at me about the modulation Transfer function yesterday, which is In some sense the part of my lecture So I found in my several Lectures I Have in my computer part of one lecture and I just tell you a few words about That just to understand what is modulation transfer Function, but before I will tell you that please remember that modulation transfer function in learning in teaching is to be in time at all lectures So if you have If you want to have a good modulation transfer function, please be in time Okay That's a professor. I say to you Okay, and now please consider such an example We have cobalt unit. This is the source quite big. It's two centimeter diameter source And we have here the block Okay, and we take a picture of that block What we will see if we draw the profile of That picture how that profile will Look like it was Well, if that will be something like this why because of the diameter of the source And that's exactly modulation transfer function if the source The source which gives you an image would be ideal then Your image would be like this okay But it's not like this because the source you used for imagination for making images Is large one. So your modulation transfer function Is not as good as You wish to have that's the first well feeling about modulation transfer function and now Let's imagine that you have periodical object, okay Well, we can say sinusoidal object. Let's imagine that you build such an object so you're just using your computer you calculate Thicknesses along the distance. What should be thicknesses to add to have attenuation Exactly equal to sinusoidal function. Okay, so your object can be Described by that function and this is the formula you can use for that Okay, that's very simple Snuffing complicated, okay, let's imagine what happens if you look at an image of that object First of all, please remember that we can describe with that formula and this is amplitude and this is average amplitude and The fact function which modulates your object, okay, and This is what you get on your image Let's imagine that that's your function and the function is described again by exactly the same function but we have different amplitude and There is Shift in face Let's imagine that there is there is usually there is okay so we have an object and we have a Detector sorry for Polish language, but I couldn't change that slide because It was an image, so It took me a problem to change it. Yeah, okay, so we have do you see the difference? The amplitude is different. Okay What is object modulation? It's does the ratio between the amplitude and the average signal Which is produced by that object. What is modulation of the detector is exactly the same What is modulation transfer function is the ratio of these two functions Okay, so modulation transfer functions by definition is this But of course that depends on the frequency If we come back to our object that was the object with a given frequency But you may have different frequencies in your object. For example the edge has inside very high frequencies and These frequencies are Not transferred because any longer you have the edge You have more smooth function so that means That Imaging system is not able to transfer very high frequencies of the object okay, and just Last slide concern modulation transfer functions. We have periodical object. We can using Fourier transform transform it to into the Frequency space so we can describe our object with different amplitudes For a given frequency. Okay, and we can look at the object which was taken with an Imaging system we can again transform it into frequency space and for each frequency we have Amplitude we have divide amplitude of the object and Image and that is Modulation transfer function. I cannot say more about that. I'm not ready for it Okay, and now I will change my presentation So many different ideas came to my mind what to say to you It's it's always when I'm listening to other lectures different ideas Come to my mind, and I think I should change a bit my presentation Okay, how to use image information IGRT. That's the second part of my talk. Okay Well, what is the most important thing? if we Consider reproducibility there is a very nice movie What is the most important thing and in that film? The most important thing is family So but we talk about reproducibility. Okay Well, not procedures not imaging systems, but good preparation a patient for radiotherapy Please remember that That that's the most important thing You I think that you don't know how the patient feels When he or she is prepared for radiotherapy How difficult is to be treated and Consequential it's difficult to cooperate with such people Do you make thermoplastic masks in your departments who do it almost everybody all everybody? Okay? Have you ever made yourself that mask who made it? Okay, and that's a very good idea Because then you can feel what a patient Feels when he or she is prepared for radiotherapy because your radiation technologies and Also you you have cooperate very well with them You should talk to them before you start preparation of your treatment because reproducibility depends on the elite of reproducibility immobilization systems Independs mostly on the behavior of a patient. So for example, it's impossible to immobilize patient in the pelvic region just impossible But if patient leaves on the table and Doesn't move that means Reproducibility so you should cooperate you find the the good position for for him or for her Which is just reproducible just try to lay on the table without having something Below your your knee then immediately you will see that it's difficult to line that position so your work Your aim is to find the best position of your patient Might be not your but you should come when you can you should talk to your radiation technologies to to explain them how it's important Here you have there these two pictures are from my department how The work can be done in not proper way you may see here This is from simulation and This is from treatment. You may see the difference in the position of the spinal cord But that is certainly because the first position was wrong that patient was was pushed to keep the Let's say artificial position for him or for her of course during the treatment here she Changed his position to the more more comfortable. So again, that's the what I think the most important message I want to to send to you Please work with your patient to prepare him or her To find the comfortable position and to explain him or her what will go on Okay, so that's that's the most important. Okay now Let's move to the to the more strict part of my presentation I'm sure that Well, you know what is tumor control probability is just probability of control of of kill all cells in the irradiated region, okay, you more control probability can be in Approximate manner described with this function. This is TCP what you would receive if Homogeneous mean dose would be delivered to to the target and this is you have to subtract from your TCP for D mean and And sigma is just standard deviation of those distribution in the clinical target volume Gamma is so-called normalize dose gradient. It tells us how much you TCP is changed by if you change the dose delivered to the patient of 1% Okay, and that fact and let's imagine that your patient for fully homogeneous dose distribution TCP is 0.5 then if Sigma increases TCP Decreases depending on the gamma if gamma is freedom. It's blue line if gamma is five is I Don't know the color the pink In general I have problems with colors man usually have some problems with colors Okay well, so if we think about Well procedures we can apply for making Our treatment more reproducible. We should think about two two terms about the mean Sorry about the mean dose so that's the dose prescribed by the doctor and Total uncertainty of radiotherapy procedure in terms of mean dose depends on the uncertainty of preparations. I mean the TPS calculation model uncertainty of realization and standard deviation is Uniformity of those distribution in the PTV and mostly it's related to reproducibility of treatment plan So if your treatment is not reproducible So the patient move during irradiation Sigma increases if Sigma increases TCP Decreases but the most important is of course mean dose if you don't deliver the right mean dose you change your TCP of a large Okay, so What that means that means that if we think about Systematic and random error systematic Errors are those error who are present all the time during the treatment. Okay, and random they change They they change from the day to Today, okay, and of course systematic error is most More dangerous because it influence on the mean dose delivered to To a patient the random error is less Important it's important, but it's less important because it influence on Sigma on Homogeneity of dose distribution in the target unless it's very big Random error then also the the the mean value is changed. Okay, so How we verify radiotherapy in space of dose you compare Prescribes those and deliver those for example with in vivo dosimeter different systems We measure before we start treatment so on and in space of location, which is portal control, which is all Things we are talking about Again we compare we have reference image and we have actual women we compare them and we find the We find the difference between them and we know what is an error error of Of the treatment again, we can use simulator for that But however, I don't recommend to use simulator or recommend use you to use digitally reconstructed radiographs. However However, right now simulators disappear from Radiotherapy department, but in my opinion, it's a mistake why Because as I mentioned to you the most important thing is to to to prepare a patient properly for your treatment How you may prepare your patients for the treatment? Of course, you may talk to a patient But Simulator is the perfect tool to prepare a patient for treatment and in my department There is a an ingenious radiation technologies who introduce so-called introductory simulation for some patients before patient starts His or her treatment He or she okay, I will tell he doesn't matter Just to make it short. Okay He comes to the simulator He's placed several times on the simulator. So the radiation technologies Place him on the table of the of the simulator take an image or images he Go out of this of the table. He they just put him again On the table they take images and then compare these images and if they see that images are very similar Each one of these images are very each one to each other. Okay Then they are sure that the position they found is right for that patient is reproducible, of course, I know that's not always it's possible because time is limited There is no stuff to make it but it's remember that it's for at least for some patients. It's a very Good procedure to prepare a patient for your treatment. Okay So this is digitally reconstructed image how we get it. We just calculated the attenuation of of the Radiation across a patient and in that way we create an image and that image should be used as a reference one So of course we we you may see the difference in quality between images This is digitally reconstructed radiograph and this is Simulator image if you want to have a very good digitally reconstructed radiograph you should take your city Cross sections every let's say two millimeters Then your your DRR will be really a very good one Okay Bow should be very narrow, you know, so two millimeters step and two millimeter reconstruction Yeah, yeah, but the problem is that if you send so many images to your doctors They don't like it very much because they have to draw so many contours so in our hospital, we Have a procedure that we make every two millimeters. We reconstruct Because that's Helico city. So that depends on how you reconstruct your images. So we reconstruct them to Every two millimeters, but we have another set of images every let's say three or four millimeters depending on the case, but Please be very careful with that because it's very easy to make a mistake Because your DRR's are Added to your plan manual So it's very easy to miss the set of city data and we had such a situation So now there is a discussion in our department. Maybe we should come back to the First idea to have exactly the same step and same Well, the same set of images should be used for the preparation of digitally reconstructed regular graphs and for drawing Contours just our problem not your problem Well, I'm just mentioning about that because I think it's it's practical issue okay, so we get portal images and You know that image But yesterday I made a mistake and I because I Didn't use the power point presentation. I use PDF file and I wanted to ask you what's that? Yes, then you've seen that zebras. Yes, but it's not easy to understand yet so well Just to show you how important are Contour sorry for that. That was my mistake. So Okay, so eventually we we got that information we have the well The vector which describes the error we made Placing a patient the setup error. Okay, and of course we may also have the angle of Rotation and again for angle of rotation if you see the angle of rotation error You cannot do anything. There are some very spatial tables that can also rotate But to be honest, I don't like it because try to put your patient on the table which is Stilted then immediately patient is trying to To come back to the previous position So I think there is no other way to correct his or her position his position Just placing him again on the table that's tedious, but anyway, you cannot do any anything different from that and Now correction strategies and that's the very important part of my talk How to correct because you collect the data you make your profile images You look at them and say it's good. It's bad and so on But you should use the data if you Don't use the data. It's just for nothing. You work so hard and there is no advantage of your work Okay, so what to do with the data? So let's remember that we try to explain our space in terms of systematic and random errors Quantitatively, what doesn't mean systematic and random errors? Let's imagine you you make the portal control every day during the treatment. So there were 25 fractions and you got Results each Diamond describes your error. So the first day at AP direction your error was close to two millimeters on fourth, you know, it's Doesn't matter fourth fraction your error was four millimeter and so on. What is systematic error? Systematic error is just the mean value of all of all these Values, okay, but you may ask well systematic error is known After completion of the treatment, that's right And that's worthless because if we know what happened It's nice to know that but it would be much better to know what Will happen? Okay Okay, we come back to that problem and the second one standard deviation is just random errors So every day we we have small difference between actual position and the mean position and if we calculate the standard deviation that describes the Random error, okay Let's imagine that we have several patients patient number one number two number three but what What are these patients? About whom I'm talking about Well It's should be the homogeneous group of patients. What doesn't be homogeneous group of a patient Treated in the same location with the same technique with the same immobilization devices and not very thin very thick or I don't know might be having some problems with spine or well Different factors that can really influence on the position of of him. Okay You should collect such a data in homogeneous groups. So we have three patients. You have three Systematic errors and you have three The random errors, okay Well at one image you have two Directions, so we have information along two perpendicular Exes so we can present that in in that form So here you have the systematic error for the num for the pink patient For the for the red one. That's the vector describing the systematic error and so on and of course the You may have many of of such patients and now mathematics first of all if You have such a group of patients homogeneous group of patients Then let's say 20 patients. So you have 20 systematic errors for your patients What would be the value of? mean over the means Zero why? But what is the reason they are canceling each other? Yeah But always we will have zero when we Don't have zero because it may happen. It's not you Okay, let's imagine. What's right now the time on your watch on my 1028 We have a very good what is so, you know, but okay, let's imagine you that you receive all your watches from One shop and in that shop There is a mistake there is an error they thought that it's ten, but it's really it's nine forty five Okay, nine forty five So the average value is not ten It's nine. It's close to nine forty five Because there is in that system there is systematic error and it can be that there is a systematic error in your whole chain for example your laser system at One accelerator is different from the another accelerator one accelerator really is well matched to the iso center But on the other accelerator is moved. Let's say three millimeters down So if you treat the patient one at the first accelerator everything is right, but if you treat them on the second accelerator Because there is a difference between city simulator or between simulator and accelerator Your systematic error will be would be three million meters So that's the most important information you get from your Portal control Collect the data for homogeneous group of patients Calculate the mean error for each of these patients and later on calculate the mean over means okay, and That value should be very close to Zero you may use statistical tests to answer the question it's It's the reason to say that it's no zero or it is zero, okay Again, that's that's really very important in our department. We have a problem We found location which is the breast location and we see the systematic group systematic error of about two to three millimeters and We fight against that problem for several years and this year we probably found the reason for that professor Hartman told you that The what now gating is used in radiotherapy But gating is not the good solution for lung patients Because they they have some problems with briefing But it's not only the case if we think about lung patients. We noticed that When the patient is placed on the table a lady And it briefs very smoothly When the table starts Just to make city examination suddenly She changes her briefing It's Longer and less deeper But dead data are used for planning but later on when she's irradiated At the accelerator the table doesn't move So it's it briefs exactly in the same way so it's very likely that when we Take the SSD from that position is different from the SSD we got from From the treatment limbic system Can you imagine two to three millimeters different we found in that way But just because we compare our data the mean data collected for portal control and for Well what we wanted to have okay our our our aim so remember that's they're very important collect the data and analyze them And very quickly because I see that my time is Coming to the end well This is also important value why I will tell you later on distribution of systematic errors So it's now it becomes more complicated. So you have homogeneous group You have systematic error for each individual patient So it's the mean value. Okay now you calculate the mean over means That should be close to zero and you calculate the standard deviation of these means and that is the Distribution of systematic errors and that's important because that value you will use to calculate the margin okay and Random group error, which is just mean value of the standard deviations obtained for each single patient Okay, let's look at that. This is this distribution of systematic error and distribution of Random under or both of them are very close to sorry This is distribution of systematic error Sorry, there is a mistake in that graph. I've never noticed that why it's a mistake Sorry for that Because distribution of systematic error can be can be minus at plus but but random error cannot be Smaller than zero there. Well, something is wrong. You that the story for that It's it's taken from Ben Heimann professor from Rotterdam, but Sorry, I don't want to criticize him because he's very wise name of liking fair match It's here's engineer medical physics is so but might be the description is wrong. So Okay strategies We have online protocols and offline protocols. What is online protocols? measure during Measure and correct in same fraction. So you take pictures you compare them and just correct the position of the patient and It seems that the best way how to correct the position of your patient Yes, but not I don't fully agree with that First of all, if you do that immediately from psychological point of view You stop working At the high level Because if you know that you can't correct Your error Well, your word become not very at very good Made with a very good very good quality And of course, it's a big effort to make everyday images to compare into irradiated patients You make your treatment longer that also deteriorates your treatment because as the longer patient Least on on the table Of course, there is a higher chance to to to misalignment between what you want and what you get, okay? so I think that offline protocols are very good and The protocol I'm I'm going to talk is no action level protocol, which is in my opinion the best protocol you can use so online Well, I just keep that slide because it's That's the online corrections. So you get image you find the discrepancy That's the question the organization of the department We're in my department in my department online correction is done by by radiation technologies and offline corrections are also made by radiation Technologists because we have ingenious group of radiation Technologies, however, they cooperate very strongly. Let's say with me I'm responsible for For all these procedures. So there is the man I mentioned about Is that man and he's really responsible for that and we prepare together all these protocols and This group is responsible for that and I can tell you that in just looking at images and to matching they're much better than me So they have big experience With that but that depends on the organization a Certainly online corrections should be made by by radiation technologies because it's difficult to to have Physicists all the time at accelerator and to have doctors all the time At the accelerator and it's not needed. I think it's not very complicated. That's if you train them well Then yeah, that's possible to do for for them. I Can imagine there is some politics behind that question, but anyway Okay, so that's the online correction Well, whatever images you use for that you can you can apply online correction and No action level the protocol so fraction first second and third Set up a patient according to protocol Make a portal control and before fourth fraction calculate the systematic error. Well, I should well Before I I Show you that slide I should make another comment. I Told you that we know the systematic error after completion of the treatment and that's right but If we assume That errors have normal distribution Okay Women estimate very quickly. Let's say the mean error. What is the best estimate of the? Mean value if the dose distribution is normal is the mean over a several Measurements, okay, you do it every day if you measure those rate you output factor you measure several times and you represent your output factors with demand value of these for Measurements, but of course you may make them a lot many of them But you think that if you if the that's the normal distribution and it's not very wide You may use you may take only a few measurements to To estimate the the mean value and that's exactly the same So we take four portal. We make four three portal controls We calculate the mean value and we say this is our estimate of the systematic error, okay So before for fraction we calculate the systematic error from for fraction on Set up a patient according to protocol. It should be protocol shift couch with the vector Okay, it's minus All these value and irradiate and that's very simple and very efficient Protocol how to make your treatment very reproducible. Okay I Waited for that question Well, that's the original no action level protocol proposed by Rotterdam group What we do in our department? We have so-called action levels so if the difference between the actual position and The position we wanted to have is smaller than let's say for the head and neck two millimeters We just leave a patient on the table, but it is three millimeters four millimeters. We shift the table during the first second and third Fraction but again our protocol is more complicated if that difference is larger than four millimeter. We stop the treatment We try to find the reason for that that we got such big in head and neck of course In pelvis we have different for example for gun gynecological Treatment these action levels are if I remember well, it's it's four eight and twelve millimeters so That makes the procedure more complicated, but it's we follow that that procedure but original No action level protocol is as I told you. Okay, so that's that's very simple after Third fraction we calculate the mean value, which is estimate of the systematic error and later on when we got result we move We shift the table of that value. Okay so the blue one is how the patient will be Irradiated and as you see the final Systematic error is very small one. Of course. There is always some residual error because That's that's clear from physics from mathematics at some residual error left Which is of that of that range sigma divided by number of Images number of results you've got to calculate the systematic error. Okay So this is what you get Without correction without protocol and what is You get with with the protocol so it's much much better and you may see it here Residual after no action level protocol if it's applied Displacements for different Locations see that for example in prostate the residual error is Very close to one millimeter. I even cannot imagine that we can irradiate our patient with such a precision That the residual error is of the range of one two millimeters. It's very precise Treatment. Okay, that I commented already. So I'm not going to talk about to talk about that So remember there's some errors in those the delivery are Inevitable so you cannot you know, you're never life is perfect. So You will make some errors and Sometimes it happens How to ensure that full prescribed dose will be delivered by adding margins margins, you know the margin between CTV and And PTV. Okay, think our target volume and planning target target volume how to calculate We will we have two type of margins setup margin internal margin setup margin is just to compensate setup errors and Internal my dreams to compensate movement of the target caused by physiology. That's much complicated problem. Yeah, okay. I'm Finishing so you calculate the systematic error random error and We have two formulas to to add margins. The first one is is Proposed by Marcel van Herk and the second one proposed by Strom They are a bit different Wow, there is a mistake here. Sorry. It's That is Marcel van Herk Formula and that is strong formula. Sorry for that I don't know how it happened, but I I recognize that I change it in my presentation. So Okay, and Now just Well, the last very important message to you The data must be collected and Regularly analyzed It's never ending story. You should all the time collect the data of portal control and you should analyze that And big errors must be analyzed as quickly as possible a conclusion must be drawn Don't leave it and don't leave it to radiation technologies. You have to do that You are able to analyze the data. Okay group systematic errors. I I I already talked about that and thank you very much