You are on page 1of 6

Medical Dosimetry, Vol. 25, No. 2, pp.

7176, 2000 Copyright 2000 American Association of Medical Dosimetrists Printed in the USA. All rights reserved 0958-3947/00/$see front matter

PII: S0958-3947(00)00031-5

BIOPLAN: SOFTWARE FOR THE BIOLOGICAL EVALUATION OF RADIOTHERAPY TREATMENT PLANS B. SANCHEZ-NIETO and A. E. NAHUM
Joint Department of Physics, Royal Marsden NHS Trust and the Institute of Cancer Research, Sutton, Surrey, UK
( Accepted 22 February 2000)

AbstractDistributions of absorbed dose do not provide information on the biological response of tissues (either tumor or organs at risk [OAR]) to irradiation. BIOPLAN (BIOlogical evaluation of PLANs) has been conceived and developed as a PC-based user-friendly software that allows the user to evaluate a treatment plan from the (more objective) point of view of the biological response of the irradiated tissues, and at the same time, provides exibility in the use of models and parameters. It requires information on dose-volume histograms (DVHs) and can accept a number of different formats (including DVH les from commercial treatment planning systems). BIOPLAN provides a variety of tools, such as tumor control probability (TCP) calculations (using the Poisson model), normal tissue complication probability (NTCP) calculations (using either the Lyman-Kutcher-Burman or the relative seriality models), the TCP method, DVH subtraction, plots of NTCP/TCP as a function of prescription dose, tumor and OAR dose statistics, equivalent uniform dose (EUD), individualized dose prescription, and parametric sensitivity analysis of the TCP/NTCP models employed. 2000 American Association of Medical Dosimetrists.
Key Words: Biological evaluation, TCP, NTCP, DVH, Radiotherapy.

INTRODUCTION Distributions of absorbed dose (either as dose maps or as condensed dose-volume histograms [DVHs]) do not provide any information about the biological response of the tumor and organs at risk (OAR). However, in recent years, there has been an increased interest in so-called biological modeling, and there are presently several models available on tumor control probability (TCP)13 and normal tissue complication probability (NTCP).4 8 Furthermore, concepts such as optimization of treatment plans according to biological cost functions are becoming accepted.9 However, the uncertainty in the currently available clinical data implies that there is no consensus on the most appropriate model and/or parameters. All of the major commercial treatment planning systems (TPS) provide DVH les, but only 3 of them include NTCP and TCP calculations. However, they are obviously linked to plans carried out in those particular TPSs. There is a lack of a exible and comprehensive tool allowing the user to evaluate a treatment plan (whatever planning system has been used) from the (more objective) point of view of the biological response of the irradiated tissues, and at the same time, providing exibility on the use of models and parameters. BIOPLAN (BIOlogical evaluation of PLANs) has been conceived as a PC-based user-friendly software for the biological evaluation of treatment plans. (BIOPLAN
Reprint requests to: B. Sanchez-Nieto, Joint Department of Physics, Royal Marsden NHS Trust, Institute of Cancer Research, Downs Road, Sutton, Surrey, UK. 71

stand-alone executable le is available on request [for non-commercial use] from the authors.) It requires information on DVHs, and can accept a number of different formats (from simple ASCII les, with differential or cumulative DVHs, to manufacturer-specic formats provided by commercial TPSs). BIOPLAN provides a variety of tools such as NTCP (using different models) and TCP calculations, the TCP method,10 DVH subtraction, plots of NTCP/TCP as a function of prescription dose, individualized dose prescription, equivalent uniform dose calculations (EUD),11 tumor and OAR dose statistics, etc. The user can access a library of model parameters; however, they can also be customized by the user, thus making BIOPLAN useful in testing the parametric sensitivity of the biological models in particular cases.

MATERIALS BIOPLAN was initially designed using Microsoft Visual Basic (a Windows development language) version 3.0; however, the most recent tools have been added using the VB 5.0 professional edition. It runs on a PC (486DX/66Mhz or higher processor) with the Microsoft Windows 95 (or higher) or NT operating system. BIOPLAN is a menu-driven application. It responds to keyboard or mouse events by displaying windows and dialog boxes. The results (numbers or plots) are displayed on the screen either in text or graphics windows, which can also be printed out or saved on disk.

72

Medical Dosimetry

Volume 25, Number 2, 2000

MODELS This section presents the equations that have been implemented in BIOPLAN to carry out TCP and NTCP calculations, as well as the parameter requirements that they imply. No explanation/discussion on their assumptions, approximations, or accuracy of their predictions is included. Tumor control probability (TCP) To compute the TCP, the Poisson model13,10 has been implemented in BIOPLAN. The equations for the model are as follows:
TCP
i

volume. Calculations ignoring the proliferation term are also allowed. (For example, calculations for prostate tumors could be done ignoring proliferation, as it is very slow [approximately 40 days].13,14) Normal tissue complication probability (NTCP) Two different models have been implemented in BIOPLAN: the Lyman-Kutcker-Burman model (empirical) and the relative seriality model (mechanistic). The Lyman-Kutcker-Burman (LKB) model. This model, also called the normal or probit model, calculates the probability of complication of a partially uniformly irradiated critical organ.4 The equations for the model are:
t

g i ( ) TCP( i,

(1)

TCP( i,
exp pc

i)

j
exp

TCP( i,
i

i,

Dj, vj)
1 dj i/

(2)
NTCP 1/ 2

exp( x 2/2) dx

(4)

vj

Dj

t (T 1 2 T k) TD 50 (v) gi ( ) exp
i

D TD 50 (v) m TD 50 (v) TD 50 (1) v


n

(5) (6)

)
2

(3)

The calculated TCP Eq. (1) corresponds to the TCP averaged over a population with variability in radiosensitivity (simulated as a Gaussian distribution of i values with mean and standard deviation) where a fraction g i of patients have i radiosensitivity and i gi 1. TCP( i , i ) Eq. (2) represents the TCP of a patient with radiosensitivity i and with a nonuniform tumor dose distribution given by {D j , v j }. The subscript j refers to the jth-volume (v j ) that receives a dose d j in each of the n fractions (D j d j n). The parameter is allowed to vary over the population of patients such that / is always constant. To allow cell proliferation during the course of the treatment, a nal term working in the opposite direction to cell killing has been added;12 where ln2/T d , T d is the average doubling time, T is the overall treatment time, and T k is the time at which proliferation begins after the start of the treatment. Thus, the parameters required by this model are: , , / , and c (initial clonogenic cell density), T d and T k. A standard set of model parameters (including the number of points to sample the Gaussian [100 by default]) are directly available (for the case of prostate tumors); however, the user can enter any other values. It is possible to avoid the -term and use instead an effective value (
eff) eff

where TD 50 (1) is the dose which, uniformly delivered over the whole organ, will produce a 50% chance of complications and v is the fraction of the organ irradiated uniformly. The 3 parameters in the model TD 50 (1), m, and n have been tabulated by Burman et al.15 for different organs and specied endpoints based on the clinical tolerance data by published by Emami et al.16 BIOPLAN contains a library with some of them, otherwise they can be entered by the user. To incorporate into Lymans model the more likely situation of a nonuniform irradiation of the critical organ, Kutcher and Burman5 gave a reduction scheme to reduce the DVH to an effective fractional volume uniformly irradiated to the maximum dose. This reduction model has also been incorporated into BIOPLAN to deal with nonuniform dose distributions. The Simpsom method is used by BIOPLAN to evaluate the function error in Eq. (3), with 50,000 integration steps and a lower integration limit of 10. The accuracy of the calculation with these values has been successfully tested against the result of a FORTRAN routine (S15AEF from the Numerical Algorithm Group library [NAG]). The relative seriality model. This model describing the probability of damage of normal tissue is based on binomial statistics.8 It accounts for serial and parallel architecture of the functional sub units (FSU).17 The expression is as follows:
k NTCP 1 j 1 1 NTCP(Dj )
s
1 s

, which
v
j

ignores the different dose per fraction inside the tumor

(7)

BIOPLAN q B. SANCHEZ-NIETO and A. E. NAHUM

73

which describes the response of the whole organ to an arbitrary dose distribution {D j , v j } as a function of the response of the whole organ to a homogeneous dose distribution. The number of FSU has been made to coincide with the k bins in the DVH, and s is the relative seriality factor. NTCP(D j ) can be expressed as
NTCP(D j) 2
e e
Dj 1 D50

Table 1. BIOPLAN ASCII le format


Record 1 2 3 4 5 6 7 8 to end of le Variable description Patient identication Organ identication Dose normalization value (%) XY and Z resolutions used in dose calculations Absolute organ volume (cm3) Prescription dose (Gy) Minimum, maximum, and average dose (Gy) Top of dose bin (%), volume in the bin (%) Type of variable String String Single Single Single Single Single Single

(8)

which is based on Poisson statistics to describe cell survival. The parameter is the maximum relative slope of the dose-response curve and D 50 is the whole organ uniform dose that would produce a 50% complication probability. BIOPLAN has a library of s, , and D 50 values for some critical organs/clinical endpoints, but any other values can be chosen. Linear Quadratic (LQ) correction to the physical DVH Dose heterogeneity in the organ at risk will have biological effects due to varying fraction sizes as well as total dose.18,19 None of the NTCP models given above deals with this fractionation-effect problem, while the parameters in them effectively correspond to a 2-Gy fraction. Thus, the physical DVH should ideally be previously converted to a 2-Gy-equivalent DVH before using the models. Physical doses D delivered with a dose per fraction d can be converted to the 2-Gy equivalent dose D2 by using the following expression:
D2 D ( / ( / d) 2) (9)

The basic structure of a BIOPLAN le consists of a header (from records 1 to 7) and the differential DVH from record 8 to the end of the le. The differential DVH (dDVH) is represented by the top of the dose bin (%) and the corresponding volume (%), separated by commas. The maximum number of bins has been set at 1000; they do not need to be equally spaced. The last bin should have 0 volume.

(dose, in Gy, in the rst column and volume, in cm3, in the second column); these are then transformed into BIOPLAN le format. Although BIOPLAN DVH les use the top of the interval to dene the dose bins, the center of the interval is the value used for calculations. Model parameters Regarding TCP, BIOPLAN has access to a default set of parameters intended for prostate calculations,10 which is displayed through a dialog box before each calculation. At this point, they can be either accepted by the user or re-entered (therefore TCP calculations are open to any tumor site). The prediction of complications for a certain organ necessarily implies the denition of clinical endpoints. Each endpoint is characterized by a different dose-response curve (even for the same organ). Thus, a different set of parameters must be chosen for each endpoint. As BIOPLAN has implemented 2 different models to calculate NTCP (LKB and relative seriality), 2 types of sets of parameters are available, depending on the model being used. There are pre-set values for the parameters;8,15,20,21 however, they can be modied by the user at any time. OUTPUT DATA Plan intercomparison A maximum of 11 BIOPLAN DVH les can be handled at the same time for the purpose of intercomparison. As they are picked up from the le-system control, they are automatically plotted on the screen in both differential and cumulative modes, using either percentual or absolute values for the axis (user option). These plots can then be printed. A dose statistic information window is available, on request, for each selected DVH. This information includes the volume irradiated to different percentages of

Equation 9 can be applied to each D j dose-bin of a DVH. BIOPLAN has incorporated the option to apply the equation above to correct the physical DVH for fractionation effects ( / values are freely selectable by the user). The resulting 2-Gy per fraction equivalent DVH is then used as the input DVH in the NTCP models described above. INPUT DATA The input data for BIOPLAN (i.e., for the models predicting complications and tumor control) consists of dose-volume data and the model parameters. Dose-volume histograms (DVHs) BIOPLAN handles DVH les (1 per structure) in a simple ASCII format (see Table 1 for description). However, it also incorporates routines to transform DVH les generated by several commercial TPSs (Target, Cadplan, and Helax) into the BIOPLAN format. This transformation starts by splitting the information contained in the original DVH le into separate les, 1 for each of the structures contained in the source le. The dose-volume information is then reorganized according to the BIOPLAN format and stored on disc. BIOPLAN can also read simple ASCII les containing a cumulative DVH

74

Medical Dosimetry

Volume 25, Number 2, 2000

the prescription dose (e.g., 50%, 60%, 80%, 90%, etc.), minimum, maximum, and average dose (with standard deviation). TCP (with/without cell proliferation) and NTCP calculations (using either of the 2 models available) can be carried out on each of the selected les, using either the pre-set parameter values or any others. Thus, the parametric sensitivity of the models can be evaluated (although another more specic tool for the case of TCP has been explicitly set up, see below). In this way, it is possible to test, for example, if the ranking of the plans (i.e., assuming that they have been scored according to either TCP or NTCP values) is affected by varying the parameters within clinically reasonable intervals. There is a window where all relevant information concerning a TCP or NTCP calculation (i.e., DVH le and parameters being used, as well as result of the calculation) is accumulated (unless deleted by the user) to help in the comparison process. BIOPLAN offers also the possibility of calculating the TCP for the structures dened by 2 separate DVHs (e.g., prostate gland plus seminal vesicles). This has the advantage of being able to choose different parameters for each of the substructures (e.g., different clonogenic cell densities). Another tool allows the user to subtract one DVH from another. The result can be stored as a BIOPLAN le and thus can be accessed in the future to perform calculations. This may be of particular importance in the case of hollow structures, such as the rectum, for which the wall is the biologically relevant structure (e.g., the DVH for the wall [DWH] can be calculated by subtracting the DVH for the inside of the organ from the whole organ). Finally, BIOPLAN can calculate the mean DVH of a user-selected group of DVHs. Again, the resulting mean DVH can be stored as a BIOPLAN le. This operation can be useful for calculating a DVH representative of the mean organ irradiation of a patient group. The contribution of each DVH to the average (i.e., its weight used to calculate the mean) is left as an input parameter. We want to add a note of caution on the mean DVH calculation for calculating, for example, a DVH representative of the whole treatment of a patient for whom there is a set of available DVHs (e.g., representing daily treatment differences due to, for example, patient/organ movement). Due to the lack of spatial information in the DVHs, the resulting mean DVH may be totally unphysical, as the values at the different dose bins will be averaged, even if they correspond to different physical volumes; the correct procedure is to rst add the different dose distributions together and then compute the DVH. Biological plan evaluation There are a number of options that allow the user a more comprehensive plan evaluation from the point of view of the biological response. These functions are

Fig. 1. BIOPLAN screen window showing the plot of TCP, NTCP, and UCP vs. prescription dose for a specic radiotherapy plan. A prostate treatment plan has been chosen as an example (DVH les for the prostate, GTV, bladder, and rectum were used). The LKB model has been used for the NTCP calculations. The different colors displayed in the computer screen identify each plot and make possible the link between curves and numbers displayed on the right. As this picture is shown in black and white, arrow lines pointing to each curve have been added to identify them.

available to 1 plan at a time (i.e., the screen is refreshed as a new tumor DVH le is selected). A plan comprises of the BIOPLAN les for the tumor volume (Gross Tumor Volume, Clinical Target Volume, or Planning Target Volume as decided), as well as for the critical organs (no number restriction). Under this single plan modality, it is possible to plot on the screen the TCP, NTCP (1 for each OAR as well as the total NTCP), and the uncomplicated tumor control probability (UCP) (UCP is computed according to the following expression: UCP TCP [1 NTCPTot ]
no. of organs

TCP

f 1

[1

NTCPf ]) as a function of the prescrip-

tion dose (see Fig. 1). By displacing a vertical line on the plot, the coordinates of the intersecting points on the curves (TCP, NTCPtot, UCP, as well as prescription dose values) are displayed on the adjacent text boxes. This tool may help to analyze the impact on outcome predictions of different dose prescription levels for a given dose-volume distribution and thus, for example, to evaluate the consequences of dose escalation. As mentioned above, there is a tool to further analyze the impact of parameters in the TCP model. Specifically, it is possible to plot, for a given tumor dose distribution, the TCP curve vs. prescription dose for a selectable range of radiosensitivities ( ), varying for each one the values (see Fig. 2). Another tool that is available is the computation of the TCP values.10 This novel method of evaluating a treatment plan consists of calculating (for each single bin in the target DVH) the gain or loss in the overall TCP as

BIOPLAN q B. SANCHEZ-NIETO and A. E. NAHUM

75

Fig. 2. BIOPLAN screen window showing the Poisson TCP model dependence on (mean of radiosensitivity over the patient population) and (standard deviation of ) for a specic tumor dose distribution. Thus, the dependence of the outcome prediction on the value of these 2 parameters can be evaluated. In this case, the effect of 4 different values of radiosensitivity together with 3 different values of has been explored. On the actual computer screen, different colors identify the curves. For the purpose of clarity, the corresponding value has been added to the side of each set of curves on the hard copy.

a result of each particular bin in the DVH not receiving the prescription dose. Thus, it is possible to evaluate the impact on the outcome that hot or cold areas may have on the TCP (Fig. 3). Although the standard method of prescription is the use of a single dose (depending on tumor site) for all patients, there is evidence to suggest that individualized prescription doses may produce benet.22,23 Thus, rather than prescribing a single dose which would inevitably produce a broad range of complication probabilities over the population, the idea is to work out, based on the individual DVH, the prescription dose that would produce for each patient a certain (clinically acceptable) level of complication (e.g., NTCP 5%). BIOPLAN is able to calculate the individualized prescription dose for each patient, for treating to tolerance (user-dened NTCP level) based on the individual DVH. Furthermore, BIOPLAN can calculate the EUD,11 which is a useful value to report (as this may correlate with clinical outcome), as well as the TCP that would arise if the tumor were uniformly irradiated to the minimum, maximum, or mean dose (taken from the DVH). CONCLUSIONS BIOPLAN began as a very small project to aid in TCP and NTCP calculations. However, it proved to be so useful that it continued to grow, and more tools were added to the initial software. We then started to freely distribute it world-wide (it has been requested from 13 different countries) and this also contributed to the fact

that BIOPLAN was modied to accept DVH les in a variety of formats. The major advantage of BIOPLAN is its exibility. It allows the user to exploit the potential that DVH information may provide. It is possible to choose different models (LKB or relative seriality for NTCP calculations), make approximations to the models (e.g., TCP calculation with/without cell proliferation and with/without the 2 term) and choose model parameters. BIOPLAN allows plan intercomparison. By simply looking at the DVHs of the rival plans plotted together (in both differential and cumulative format and using either absolute or relative values), one may guess the optimal plan. However, these plots are not often conclusive, as the DVHs may cross each other. BIOPLAN provides a new insight into the problem by calculating the TCP and NTCP values associated with the rival plans. Thus, a decision can be taken according to these predictions. Even when the models/parameters have not been clinically validated, and thus the absolute value of the predictions may be not accurate, the ranking (i.e., the relative comparison) may well not be affected. It is possible to verify this by performing a parametric sensitivity analysis on the models using the tools that BIOPLAN provides. Apart from the plan intercomparison it is possible to analyze/enhance a particular treatment through the prism of the biological response. Thus, it is possible to assess,

Fig. 3. BIOPLAN screen window showing the TCP values that correspond to a selected GTV DVH le (displayed on the left-hand side). Each bin of the histogram has also been represented in the pie chart where the area of each section is proportional to the volume of the bin. The number on the top of each section is the TCP value that corresponds to that particular bin. When the mouse is moved on the pie chart, a cross over the corresponding bin on the DVH is displayed and the number in the rst, second, and fourth box, that correspond to the volume, dose, and TCP, respectively, are also refreshed. The actual TCP and the value which would arise if the prescription dose were uniformly delivered to the tumor volume are shown on the top of the DVH and in the third text box, respectively.

76

Medical Dosimetry

Volume 25, Number 2, 2000 volume method. Int. J. Radiat. Oncol. Biol. Phys. 16:16231630; 1989. Niemierko, A.; Goitein, M. Calculation of normal tissue complication probability and dose-volume histogram reduction schemes for tissues with a critical element architecture. Radiat. Oncol. 20:166 176; 1991. Niemierko, A.; Goitein, M. Modeling of normal tissue response to radiation: the critical volume model. Int. J. Radiat. Oncol. Biol. Phys. 25:135145; 1993. Kallman, P.; gren, A.; Brahme, A. Tumour and normal tissue responses to fractionated non-uniform dose delivery. Int. J. Radiat. Biol. 62:249 262; 1992. Niemierko, A.; Urie, M.; Goitein, M. Optimization of 3D radiation therapy with both physical and biological end points and constraints. Int. J. Radiat. Oncol. Biol. Phys. 23:99 108; 1992. Sanchez-Nieto, B.; Nahum, A.E. The delta-TCP method: a clinical useful measure of tumour control probability. Int. J. Radiat. Oncol. Biol. Phys. 44:369 380; 1999. Niemierko, A. Reporting and analysing dose distributions: a concept of equivalent uniform dose. Med. Phys. 24:103110; 1997. Fowler, J.F. The linear quadratic formula and progress in fractionated radiotherapy. Br. J. Radiol. 62:679 694; 1989. Fowler, J.F.; Ritter, M.A. A rationale for fractionation for slowly proliferating tumors such as prostatic adenocarcinoma. Int. J. Radiat. Oncol. Biol. Phys. 32:521529; 1995. Haustermans, K.M.G.; Hoand, I.; van Poppel, H.; et al. Cell kinetic measurements in prostate cancer. Int. J. Radiat. Oncol. Biol. Phys. 37:10671070; 1997. Burman, C.; Kutcher, G.J.; Emami, B.; et al. Fitting of normal tissue tolerance data to an analytic function. Int. J. Radiat. Oncol. Biol. Phys. 21:123135; 1991. Emami, B.; Lyman, J.; Brown, A.; et al. Tolerance of normal tissue to therapeutic irradiation. Int. J. Radiat. Oncol. Biol. Phys. 21: 109 122; 1991. Withers, H.R.; Taylor, J.M.G.; Maciejewiski, B. Treatment volume and tissue tolerance. Int. J. Radiat. Oncol. Biol. Phys. 14:751759; 1988. Withers, H.R. Biological basis of radiation therapy. In: Perez, C.A.; Brady, L.W., eds. Principles and Practice of Radiation Oncology. Philadelphia: Lippincott Company; 1992:64 96. Wheldon, T.E.; Deehan, C.; Wheldon, E.G.; et al. The linearquadratic transformation of dose-volume histograms in fractionated radiotherapy. Radiat. Oncol. 46:285295; 1998. Gagliardi, G.; Lax, I.; Ottolenghi, A.; Rutqvist, L.E. Long-term cardiac mortality after radiotherapy of breast cancerapplication of the relative seriality model. Br. J. Radiol. 69:839 846; 1996. Gagliardi G. Modeling heart and lung complication data in radiation therapy of the breast. Ph.D Thesis. Department of Medical Radiation Physics. Stockholm University, 1998. Sanchez-Nieto, B.; Nahum, A.E.; Dearnaley, D.P. The customisation of dose prescription based on individual dose-volume and radiosensitivity data. Radiother. Oncol. Submitted. MacKay, R.I.; Hendry, J.H. The modelled benets of individualising radiotherapy patients dose using cellular radiosensitivity assays with inherent variability. Radiother. Oncol. 50:6775; 1999.

for example, (a) the effect of a particular nonuniform dose distribution upon the probability of controling the tumor, i.e., by how much the TCP is decreased due to the presence of an underdosed subvolume and vice versa, (b) the effect of the dose prescription on TCP and NTCP (and specically compute the individualized prescription dose which would produce a certain level of complications), and (c) how the fractionation regime affects the outcome for a particular patient. Although BIOPLAN may play an important clinical role in the daily radiotherapy treatment planning design, it can also become a very useful research tool. It can be used to evaluate/ne-tune the models implemented by retrospectively assessing the risk of complications or the probability of tumor control for a patient population for which clinical data on outcome are available. It can also contribute to the improvement of treatment techniques according to predictions of the outcome (especially probability of radiation-induced effects). In summary, BIOPLAN is a very versatile piece of software that can improve the planning process, either by saving time or increasing the quality of the results. Furthermore, it offers tools that may help in carrying out research on biological modeling.

6.

7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

AcknowledgmentThis work has been supported by the Cancer Research Campaign (CRC).

REFERENCES
1. Nahum, A.E.; Tait, D. Maximising local control by customised dose prescription for pelvic tumours. In: Breit, A., editor. Advanced radiation therapy tumour response monitoring and treatment planning. Heidelberg: Springer-Verlag; 1992:425 431. 2. Webb, S.; Nahum, A.E. A model for calculating tumour control probability in radiotherapy including the effects of inhomogeneous distributions of dose and clonogenic cell density. Phys. Med. Biol. 38:653 666; 1993. 3. Niemierko, A.; Goitein, M. Implementation of a model for estimating tumor control probability for an inhomogeneously irradiated tumor. Radiother. Oncol. 29:140 147; 1993. 4. Lyman, J.T. Complication probabilities as assessed from dosevolume histograms. Radiat. Res. 104:S13S19; 1985. 5. Kutcher, G.J.; Burman, C. Calculation of complication probability factors for non uniform normal tissue irradiation: the effective 19. 20. 21. 22. 23.