Radiation induced lung damage - Chapter 9 Discussion

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Radiation induced lung damage

Seppenwoolde, Y.

Publication date

2002

Link to publication

Citation for published version (APA):

Seppenwoolde, Y. (2002). Radiation induced lung damage.

General rights

It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations

If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

Chapterr 9

Discussion n

Discussion n

Thee aim of the study described in this thesis was to measure and predict radiation induced pulmonaryy damage for patients with non-small cett lung cancer (NSCLC) and to use this knowledgee to improve irradiation by minimizing normal tissue complications. We determinedd (actors associated with radiation induced pulmonary damage by measuring thee effects of irradiation on local and overafl pulmonary function. We tasted dose-volume parameterss as predictors for radiation pneumonitis as weH as for perfusion damage and reductionn in overall lung function. Next to these dose-volume parameters, we used function-weightedd parameters to improve radiation treatment plans. Furthermore, we determinedd and modeled factors such as patiem sebjp inaccuracy artf tuino^

hinderr accurate dose delivery. In the first part of this chapter aspects related with normal tissuetissue damage are described. In the second part the impact of function-weighted treatmentt optimization is discussed. Finally, specific possibilities and limitations for improvingg radiotherapy for lung cancer patients are considered.

Normall tissue damage

inn order to achieve better local control, the prescribed dose to the tumor has to be increased. However,, a higher dose to the tumor will cause extra radiation induced damage to the surroundingg healthy tissue. Graded response of tung tissue to irradiation was reflected in local changess in tissue density, tung perfusion and in overall reduction of pulmonary function. We observedd radiation pneumonitis grade 2 or higher, a binary type of complication, in about 10% off the patients.

Locall dose effects

Dose-effectt relations for local changes in lung perfusion were determined using SPECT and CTT scans. In our institute we have previously investigated dose-effect relations for lung perfusionn and air-filled fraction (Theuws 1998a) for patients with healthy lung tissue (malignant lymphomaa and breast cancer patients) in the dose range 0 to 55 Gy. We could extend these dosee effect relations for perfusion and air-filled fraction to local doses up to 80 Gy because lungg cancer patients are irradiated with higher local doses (chapter 3). The presence of intra-thoracicc tumor or pre-existent lung disease did not influence the response for the parts of the lungss that were heartily prior to treatment, since the response of these parts was similar to the responsee of the lymphoma and breast cancer patients from previous studies {Marks 1997b, Theuwss 1996a; 2000, Garlpagaoglu 1999).

Reperfusion n

Nott alt parts of the lungs reacted equally to locally applied dose for perfusion (in contrast to thee air-filled fraction); some parts that were hypo-perfused before treatment showed recovery (reperfusion).. Several other groups observed reperfusion after radiotherapy as well, for examplee as measured with planar perfusion scans (Goldman 1969, Fazio 1979). Later Marks ett al. (1995) observed reperfusion in their patient group and decided to exclude patients from analysiss who had reduction of perfusion adjacent to the tumor and likely would show re-perfusionn after treatment.

Wee visualized the location and extension of re-perfusion by comparing measured and predictedd post-RT perfusion (chapter 3). Coherent regions, which showed less damage than predicted,, were mostly situated near the tumor. In many cases re-perfusion extended to other

Chapter9 Chapter9

regionss and appeared to be dose-independent. Unfortunately, the amount of reperfusion could nott be predicted accurately, because not all hypo-perfused regions showed reperfusion and wee observed only a poor correlation between pre-treatment hypo-perfusion and reperfusion. Limitationss of SPECT

Theree are many uncertainties in measuring dose dependent changes in both CT and SPECT: setupp errors, breathing artifacts, image fusion inaccuracies. By applying several filters before analysiss and by averaging effects over many patients, we tried to reduce the effects of the measurementt uncertainties, but of course some practical limitations of the technique still remain,, for example SPECT artifacts as scatter and blurring. The dose-effect relation for local perfusionn damage seems to plateau at approximately 60% damage (at 65 Gy) instead of at the expectedd 100% (Garipagaoglu 1999, Seppenwoolde 2000). Attenuation of primary photons, imagingg of both scattered and unscattered photons in the photo peak energy window and finite spatiall resolution of the used camera influence the accuracy of activity quantification in SPECT imagess (Link 1996). This means that the number of recorded SPECT counts do not always representt true radioactive concentrations. Scattered photons appear as false sources both withinn and beyond the organ, resulting in a decline in image contrast, smoothing of the edges andd counted events occurring in regions were no activity is to be found. Due to the finite thicknesss of the collimator of the gamma camera, a point source at a certain distance will not onlyy be detected by just one cavity in the camera; some neighboring photo-multipliers will also detectt intensity coming from tire point source, also causing blurring of the image. When superimposingg correlated CT-defined lung contours on SPECT lung perfusion scans, we observedd that on average 30% of the SPECT counts are detected beyond the CT-defined lung.. Simulations on the effect for patients irradiated with mantle fields, using ideal and blurred images,, did not affect the dose-effect relation for local doses up to 55 Gy. However, when smallerr volumes are irradiated with steeper dose gradients, scatter and blurring can influence thee measured dose-effect relations. This might lead to an underestimation of the degree of regionall injury and to an overestimation of the reperfusion in the higher dose bins. The plateauingg of the dose-effect relation at local doses of more than 65 Gy suggests that this mightt be the maximum reduction in local function that can be identified with SPECT. Drawing aa sigmoid curve through the date and resetting the limits at 0 and 100% can then be consideredd to be a reasonable estimate of the dose-effect relation for local lung injury (Garipagaogluu 1999).

Pulmonaryy function

Recoveryy due to reperfusion was not reflected in the overall lung function of the patients as measuredd with classical pulmonary function tests (chapter 4). This can imply that, although recoveryy of perfusion after irradiation occurs; the membrane that takes care of the oxygen transportt is probably damaged permanently. Tumor regression was only correlated with reperfusionn for central tumors. We found that for patients with NSCLC, perfusion weighted dose-volumee parameters correlated weakly but significant with the reduction in pulmonary functionn in contrast to the pure dose-volume parameters like the mean lung dose.

Ann imperfection of pulmonary function teste is the variation in the measurements: a scatter of 10%% in the measurement outcome is reported for healthy persons. The system is standardized inn a way that presumes homogeneous lung function and gives even less reliable results in lung cancerr patients whoo often have difficulties performing the tests as well. In these patients pre-ëxistentt pulmonary disease and smoking further increases the variation in repeated

Discussion Discussion

measurements.. Since there is no standardization for these factors, the targe errors in the measurementss of the pulmonary function tests complicated statistical analysis.

Radiationn pneumonitis

Too estimate the incidence of radiation pneumonitis we fitted the parameters of different models too the measured incidence of radiation pneumonitis for 382 patients. In this analysis (chapter 5)) we used the full lung DVHs of the patients. All different models converged to a linear local dose-effectt that equals the approach of the mean lung dose model. The threshold dose model provedd to be significantly worse than the mean lung dose. However, due to the high correlation off the mean lung dose and a threshold dose in the studied patient group, both simple parameterss can be used for prediction of radiation pneumonitis, as long as the dose distributionss are similar to the studied ones. For predictions based on DVHs that do not belong too the class of DVHs that were studied, a model that takes into account the shape of the DVH (i.e.. the LKB model) in the calculation of the uncertainty in the NTCP should be used. With this model,, an NTCP value, calculated for a DVH with a shape similar to the DVHs of the studied patientss wilt have less uncertainty than the same NTCP calculated from a DVH with an unusual shape. .

Thee parameters of the models tested in this thesis agreed with parameters found in other studiess for the mean lung dose and the seriality model (Kwa 1998a, Gagliardi 2000), The parameterss for the threshold dose model were somewhat different compared to a study with fewerr patiënte and a less detailed model (Graham 1999).

Perfuston-wetghtedd parameters

Untill now only 'metrics' (dose volume parameters) were used for the prediction of the NTCP forr radiation pneumonitis. Because function (perfusion) weighted parameters as the mean perfusionn weighted lung dose are more predictive for reduction in pulmonary function than the puree dose-factors (chapter 4), future studies are designed to test if function weighted DVHs cann also be used for a more accurate estimation of the NTCP for radiation pneumonitis for individuall patients.

Treatmentt optimization

Noww that the mechanisms of radiation induced local and overall pulmonary damage can be modelled,, these results were used to improve radiotherapy.

Locall lung function to optimize treatment ptans

Contraryy to the expectations, optimization on the mean perfusion weighted lung dose did not resultt in different treatment plans for most classes of perfusion defects because we found that thesee defects were in general situated "downstream" of the tumor, in the parts of the lungs that weree irradiated by the beam with the shortest path through lung tissue (chapter 6). However, patientss with one hypo-perfused hemi-thorax benefit from AP-PA irradiation of the ipsMateral lungg because this reduces dose in the well-perfused contra-lateral lung. To select patients that benefitt from AP-PA irradiation, planar perfusion scans suffice instead of full 3D SPECT perfusionn scans. Advanced techniques such as multi-segment IMRT and non-coplanar beam incidencee angles might potentially improve treatment plans based on perfusion-weighted parameters. .

ChapterQ ChapterQ

Setup p

Systematicc errors are mainly due to errors that occur during the preparation phase of the treatment.. These can be due several factors such as target delineation errors, position of the patientt and the tumor when the planning CT scan is acquired compared to that during treatmentt or the use of simulator images rather than DRRs for setup reference images (Bel 1994,, de Boer 2001). Random errors can be due to day-to-day variations in patient position andd organ motion. As discussed by van Herk (2000), the impact of systematic errors on the dosee received by the CTV is far greater than the impact of random errors. The off-line correctionn protocol used at the Netherlands Cancer Institute corrects tor systematic but not tor randomm errors. Using the protocol only 1% of the patients have a systematic setup error of moree than 5 mm vector length, compared to 41% if the protocol would not have been used (chapterr 7). To improve the setup accuracy, an orthovott system can be installed in the treatmentt room. With such a system, two orthogonal images can be made each day, and the patientt position can be corrected on-line, thereby reducing systematic and random errors. Tumorr motion

Wee measured tumor motion using two techniques, first retrospectively with electronic portal imagingg (chapter 7). The tumor motion could not be measured in three directions simultaneouslyy with this method and ft was hampered by minor delineation and matching problems.. Later we measured tumor motion with real-time tumor tracking (chapter 8). We observedd that tumors that were situated in lower lobes moved more in the cranial-caudal directionn than upper lobe tumors. With the real-time tumor tracking system, using an implanted gok)) marker, we could study tumor motion in much more detail, resulting in data about hysteresiss and heartbeat induced tumor motion. The measured tumor motion reported in this thesiss agrees with other studies on tumor motion (Ross 1990, Baiter 1996, Samson 1999, Chenn 2001a).

HowHow can foiowfedge about tumor motion be used?

Thee detailed measurements reported in chapters 7 and 8 could be used to model breathing inducedd tumor motion. The data could be used to calculate the effect of breathing motion on thee cumulative dose-distribution in the CTV during the course of treatment. When a part of the CTVV moves outside the 95% isodose level, the reduction of the dose in tills part is small due too the shallow penumbra in lung tissue. In addition, the low dose will be compensated when thiss part of the CTV moves into a high dose region. The same applies for random setup errors iff the number of fractions is large enough. The effect of breathing and random setup errors on thee dose distribution in the CTV is small compared to that of systematic errors, even for the largestt breathing amplitude (Engelsman 2001a). However, these simulations did not take into accountt hysteresis. Using conventional field sizes and homogeneous target irradiation, an extraa margin for breathing motion is not necessary, especially for the upper lobe tumors and tumorss attached to bony structures because for these tumors breathing induced tumor motion iss very small. However, for tumors that move a lot more than 20 mm or when IMRT treatments withh small segments are given and the beam penumbra is sharpened, breathing motion should bee monitored and controlled (Engelsman 2001a).

Representativee tumor position

Inn the study described in chapter 7 we observed that for a number of patients the technique of performingg planning CT scans during free breathing might produce a substantial error, reflectedd in the difference between the tumor position in the planning CT compared to the averagee tumor position during treatment. There can be several causes for this difference; free

Discussion Discussion

breathingg CT scanning causes a distortion of the tumor shape (Shimizu 2000) and an unknown shiftt in tumor position. Furthermore, the measured average tumor position during treatment mayy have contained errors because of the limited number of EPIDs during one fraction or changess in breathing level. Extensive research on this subject has to be done because a systematicc error, caused by a non-representative CT scan will have a large impact on the dose distributionn in the tumor. There are several techniques to monitor or control tumor motion and too produce a CT scan that gives the best representation of the tumor:

ABC ABC

Activee Breathing Control (ABC) uses a flow monitor that measures the volume of inhaled and exhaledd air. At a pre-set point in the breathing cycle, the airflow of the patient can be temporarilyy blocked by a scissor valve, thereby immobilizing breathing. Reproducible tumor positionss during repeated breath-holds can be achieved with this method (Wong 1999). However,, a disadvantage of this technique is the breathing resistance of the tube, resulting in ann unnatural breathing level, even when the dead space of the device was considerably reduced.. Preliminary measurements revealed that the tumor position in the free breathing planningg CT scan was different than the tumor position in the free breathing ABC scan (Figure 1).. This is probably caused by changes in breathing level caused by breathing through a tube. Thiss means that to ensure similar tumor positions during scanning and treatment, the ABC devicee should be used both during CT scanning and during radiotherapy. Because a majority off the lung cancer patients often have impaired lung function, repeated long breath-holds are hardd to achieve and endure. This presents a clinical dilemma, as these are often the patients inn whom the volume of normal lung irradiated should be kept to a minimum. The use of multiple shortt breath holds may decrease treatment or scanning efficiency considerably, in particular whenn several beams with different segments are to be used.

FigureFigure 1. Overlaid sagital images of a planning CT scan (upper left and lower right) and an ABC scan (upper right and lowerlower left), both during free breathing, matched on the bony structure of the spinal cord. Note the lower position of both

ChapterChapter 9

thethe tumor and diaphragm in the ABC scan compared to the planning CT scan. VoluntaryVoluntary breath-hold

Thee problems of differences in breathing level, induced by the dead space of an ABC device cann be overcome by asking the patient to hold his or her breath voluntarily a t for example, mid-inspiration.. This approach requires thorough training of the patient and lacks the possibilityy of monitoring and controlling the breathing level in case the patient unintentionally startss breathing.

DeepDeep inspiration

Deepp inspiration breath-hold (DIBH) is sometimes used because of dosimetric considerations (Hanleyy 1999, Man 2000, Rosenzweig 2000, Kim 2001). It may be favorable to increase the out-of-fieldd lung volume to reduce the mean lung dose; for lung cancer, a significant reduction inn calculated NTCP of the lung was seen when comparing treatment plans based on CT data madee during maximum inhalation and plans based on CT data gathered during maximum exhalation.. The reproducibility of thé DIBH level is in some studies monitored with ABC or usingg external markers, but in other studies the patients hold their breath voluntarily and unmonitored,, compromising treatment accuracy.

Real-timeReal-time tumor tracking

Duringg real-time tumor tracking radiotherapy, the 3D position of an implanted gold marker is detectedd by a fluoroscopy system that is installed in the treatment room. A linear accelerator iss triggered to irradiate the tumor only when the gold marker is located within a certain region (Shiratoo 1999; 2000b, Shimizu 2001). The system provides on-line tracking of the position of thee tumor. Radiation can be delivered very precisely, provided that the implanted marker does nott migrate and the treatment planning is performed relative to the position of the gold marker andd the tumor.

Althoughh the real time tumor tracking system is the most elegant of the techniques, it is an expensivee and invasive way to control tumor motion. Because of the low impact of breaming motionn on the actual delivered dose in the tumor (Engelsman 2001a), it will suffice to concentratee on reducing the systematic error in tumor position caused by CT scanning. The ideall breathing phase in which the tumor has to be scanned should be representative for the timee averaged tumor position.

Positronn emission tomography

Treatmentt planning is normally based on clinical and detailed anatomical information that is providedd by CT scans. CT gives 3D information on tissue densities in the patient and has the advantagee that the images can be directly used for dose calculations. Although the resolution andd contrast of the images are of high quality, sometimes distinction between tissues with equall density but with different anatomic or metabolic functions is hard to make. In these cases positronn emission tomography (PET) with a tracer that indicates metabolic activity, F-18 fluorodeoxyglucosee (18FDG), will provide 3D information of regions that have a high uptake of thee tracer, like radiolabeled glucose in tumors.

Mediastinall lymph node staging

Computedd tomography has long been the most important modality in noninvasive clinical stagingg of NSCLC, but is at times inadequate, particularly in the evaluation of mediastinal lymphh nodes or distant metastasis (Webb 1991, McLoud 1992). Higher sensitivity and specificityy are reported for PET staging in potentially resectable NSCLC. Clinical-pathologic

Discussion Discussion

correlationn studies (Pieterman 2000, Scott 2001, Vansteenkiste 1998, Vanuytsel 2000), includingg a meta-analysis (Dwamena 1999), have shown that PET or a combination of PET andd CT is more accurate than conventional imaging in the detection of mediastinal metastases inn patients wrth NSCLC. PET scanning frequency can detect unsuspected distant metastases (Erasmuss 1997, Weder 1998, Bury 1993). PET scanning often changed or influenced managementt decisions in patients with NSCLC (Mac Manus 2001, Katff 2001). Patients were frequentlyy saved unnecessary treatment, and management was more appropriately targeted. Alsoo our own prospective data show that PET scan findings strongly influence the target delineation:: in 6 of 36 patients CT negative lymph nodes were found to be active on PET and inn 8 patients CT positive lymph nodes were not active on the PET scan. Currently, onty PET positivee nodes are included in the planning target volume because of the higher specificity and sensitivityy of PET concerning lymph node staging at the NKI.

Tumorr delineation

Microscopicc extension of the disease

Onee of the most difficult steps of three-dimensional conformal radiotherapy is to define the clinicall target volume according to the degree of local microscopic extension. Microscopic extensionn was found to be different between adenocarcinoma and squamous cell carcinoma (Giraudd 2000). in that study, the usual clinical target volume (CTV) margin of 5 mm appeared inadequatee to cover the microscopic extension for either group, and must be increased to 8 mmm for adenocarcinoma and to 6 mm for squamous ceil carcinoma to cover 95% of the microscopicc extension. In a second study the delineation on the CT scan was compared wrth thee size of the actual resected tumor. The findings in this small study (Chan 2001) in non-small ceill lung cancers are interesting and provoking: there was an obvious trend that the CT-gross tumorr volume (GTV) was bigger than, or equal to the resected GTV. Further study with a larger numberr of patients and more rigid quality control (especially concerning eventual tumor shrinkagee during pathology preparations) is warranted to confirm these findings.

Tumorr shrinkage

Duringg the course of radiotherapy, the apparent size of the tumor may alter due to re-opening off atelectasis or because of tumor shrinkage. When tumor shrinkage is noticed during portal imagingg and verified by a repeat CT, radiation induced lung toxicity can be reduced by adjustingg field sizes to the smaller tumor volume or the total dose to the tumor can be escalatedd by boosting the remaining tumor volume without increasing normal tissue complicationn probability. For 40% of the patients analyzed in chapter 7, we observed that the tumorr volume dropped below 80% of the original volume in the 4th or 5* week of treatment However,, it should be noted that though the size of the visible lesion on a CT scan is smaller, itt is unknown what has occurred exactly on a microscopic level.

Interr and infra-observer variation

Itt is often difficult to determine the tumor extent on CT scans (Quint 1995). Significant inter-cliniciann differences persist in the treatment planning of lung cancer despite the use of a contouringg protocol. These differences appear to be the result of differences in contouring of thee mediastinal CTV and due to window settings used. Measures to minimize the variations andd evaluate compliance with contouring protocols should be incorporated into clinical practice (Senann 1999).

ChapterChapter 9

Delineationn of primary tumor with PET

AA current problem in the delineation of target volumes in lung cancer is the differentiation betweenn tumor and atelectasis, which is difficult using the morphologic information given by computedd tomography. These difficulties often lead to the irradiation of large volumes of possiblyy healthy lung tissue. PET has the unique ability to visualize glucose uptake in vivo. Mostt tumors have an Increased glycolysis compared to normal tissue, and therefore PET has gainedd increasing interest in oncology. In lung tumors, abnormally high uptake of l6FDG was detected,, whereas in the region of atelectasis relatively low uptake of 18FDG was observed (Kawabee 1999). The experience of Duke University (McAdams 1998) also suggests that atelectasiss is not metabolically active on PET imaging. Thus, PET scans can play a role in differentiatingg atelectasis from malignant tumors if there are few or atypical features of atelectasiss on chest radiographs and CT. In a retrospective analysis (Nestle 1999) to study the potentiall contribution of PET in radiotherapy planning for lung cancer with respect to tumor-associatedd atelectasis, the information provided by PET would have contributed to a substantiall reduction of the size of radiotherapy portals. This applied particularly for patients withh tumor-associated atelectasis. Monitoring for reopening of the atelectasis during radiotherapyy by repeat CT scans, resulting in changes in the target volume, may help to preservee healthy lung tissue.

Littlee is known about PET and its accuracy with respect to size and shape of the primary tumor. Thee literature available suggests that the size may be represented quite accurately compared too CT images (Zasadny 1996), especially when using the attenuation-corrected mode of operation.. An automatic image segmentation scheme was developed to determine the tumor volumee of lung tumors from PET images (Erdi 1997). A fixed threshold value obtained from phantomm data demonstrated good correlation between the tumor volume on CT and the volumee derived from the PET scan. In patients with confirmed tumors, inflammatory disorders mayy lead to an overestimation of the tumor size: pulmonary infection or inflammation can resultt in localized 18FDG uptake mimicking pulmonary metastases and limiting the specificity (Bakheett 2000).

Dosee calculations

Thee dose calculation algorithm used in this study did not account for the increased range of secondaryy electrons in low-density lung tissue. Especially the steepness of the penumbra is overestimatedd by the current treatment planning system. Only recently, algorithms became availablee that accurately predict dose distributions in inhomogeneous media like lung tissue. Recalculationn of DVHs, dose-effect relationships and the parameters of the models used in thiss study, with accurately calculated doses, will make the results more generally applicable. Dosee escalation

Inn the current phase l/ll dose-escalation study in The Netherlands Cancer Institute the pre-determinedd toxicity levels have not yet been reached for any of ttie patient groups (Table 1). Inn the near future, a phase III study will be started to compare high dose (> 87 Gy) radiotherapy withh thé conventional treatment of 67 Gy. Because the limit in dose escalation with conventionall field sizes will be reached early for patients with large tumors, a different way of dose-escalationn is needed, for example by abandoning the homogeneous target dose. By allowingg a more inhomogeneous dose to the tumor, reduction of field sizes will lead to escalationn of the prescribed and minimum dose and an increase of the equivalent uniform dosee to the tumor, while the dose to surrounding (lung) tissue is kept constant (Engelsman 2001c). .

Discussion Discussion

Tablee 1. Radiation dose escalation steps.

Pneumonitis s >> grade 3 (max.estt %) 5.8 8 5.9 9 6.0 0 6.1 1 6.3 3 6.5 5 6.7 7 7.2 2 7.7 7 8.5 5 9.7 7 11.0 0 13.2 2 16.5 5 20.0 0 23.5 5 Groupp 1 0-0.12 2 Loww risk 74.3 3 81.0 0 87.8 8 94.5 5 101.3 3 Groupp 2 0.12-0.18 8 67,5 5

7A3 3

Subdivisionn into nek groups was based on flue mean lung dose relative to the prescribed tumorr dose. For the marónum estimate of the incidence of radiation pneumonitis £ grade 2, the Lyman-Kutcher-Burmann model is used with parameter values: TD5„*30.5 Gy, m=0.3, n=1,0 and ann offset of 11 %. Radiation pneumonitis S grade 3 was estimated to occur in 50% of the cases of gradee 2 radiation pneumonitis. The dose levels that we started with are underlined. The dose-levelss ihat we have safely reached are in bokt.

Thee relation between the equivalent uniform tumor dose and increased tumor control has not yett been validated in the clinic. The new trial will provide us with more data about local recurrencess and the 30 dose distribution in the tumor. Another way to escalate the tumor dose byy keeping dose to organs at risk low is to boost on the remains of active tumor by using repeat CTT scans.

Conclusions s

Thee work described in this thesis resulted in improved knowledge of local and overall radiation inducedd lung damage for patients with tntra-thoracic tumors, A dose-effect relation for local changess in king perfusion is derived, the perfusion weighted mean lung dose is correlated with thee reduction in pulmonary function and the mean lung dose itself can be used to predict the

incidencee of radiation pneumonitis. Using this knowledge, treatment plans for new patients can bee judged and eventually adjusted so that either radiation induced tung damage is minimized orr the prescribed dose can be escalated in a controlled manner, without increasing normal tissuee damage. This can also be achieved by reducing field sizes in the last week of treatment whenn tumor shrinkage is present The detailed information regarding tumor motion and set-up errorss can be used in future studies to improve tumor coverage.