- Email: [email protected]
A new method to determine dose-effect relations for local lung-function changes using correlated SPECT and CT data
IR ADIOTHERAPY aQ~COLOG~ Radiotherapy and Oncology 29 (1993) 110-I 16
A new method to determine dose-effect relations for local lung-function changes using correlated SPECT and CT data L. J. Boersma*a, E.M.F. Damena, R.W. de Boera, S.H. Mulleravb, R.A. ValdCs Olmosb C.A. Hoefnagelb, C.M. RoosC, N. van Zandwijkd, J.V. Lebesque” “Department of Radiotherapy, bDepartment of Nuclear Medicine, dDepartment of Pulmonary Medicine, The Netherlands Cancer Institute (Antoni van Leeuwenhoek Huis), Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands ‘Department of Pulmonary Medicine, Academic Medical Center, Meibergdreef 9, IIOS AZ Amsterdam, The Netherlands
(Received 5 Febuary 1993; revision received 3 May 1993; accepted 17 May 1993)
Abstract Purpose: To determine dose-effect relations for regional lung-function changes after radiotherapy. Methods: Single Photon Emission Computed Tomography (SPECT) was performed to quantify regional ventilation and perfusion. CT scans were used to
calculate the three-dimensional (3-D) dose distribution. Both SPECT and CT scans were performed prior to radiotherapy and 5 months after the start of the treatment. To obtain combined 3-D information on ventilation, perfusion and dose, the SPECT data were correlated with the corresponding CT data. The relative changes in ventilation and perfusion were calculated in each SPECT voxel (voxel size about 6 x 6 x 6 mm) and related to the dose in that voxel. The average relative changes were determined per dose interval of 4 Gy. This procedure was evaluated using the data from five patients treated for Hodgkin’s disease with mantle field irradiation with a prescribed total dose of 40-42 Gy. Results: Dose-effect relations for perfusion were observed in all patients, while in four of the five patients, a dose-effect relation was found for ventilation. The maximal uncertainty of the calculated radiation dose was 11%: a difference between the position of the patient during treatment and during CT scanning caused a maximal dose uncertainty of 6%, while the accuracy of the dose calculation algorithm itself was estimated to be within 5%. Conclusion: The results indicate that the combined use of SPECT and CT information is an effective method for determining dose-effect relations for regional lung function parameters in each individual patient. Key words: Radiation-induced
lung damage; Dose-effect relations; SPECT ventilation/perfusion;
1. Introduction The lung is one of the major dose-limiting organs in treatment of malignant tumors of the thorax. Radiation-induced lung damage occurs in two phases: acute radiation pneumonitis, 1-7 months after treatment, and pulmonary fibrosis, developing from 8 months onwards after radiotherapy 171. The probability and severity of the radiation damage
the radiotherapeutic
depends on radiation dose and on irradiated volume. Emami et al. [4] performed a literature study, from which the total doses with a 5% and 50% complication * Corresponding author.
probability were determined for three different volume categories, with radiation pneumonitis as endpoint (clinical symptoms and characteristic chest X-ray changes). Mah and coworkers [ 12,131 found a doseeffect relation for local damage of lung tissue, as defined by the incidence of CT density changes. To estimate the clinical toxicity, however, it is important to know not only the incidence of radiation pneumonitis or CTdensity changes, but also the functional changes, associated with irradiation of (parts of) the lung. Overall pulmonary function tests, performed within 2 years after irradiation of the lungs, showed a mixed restrictive and obstructive lung disease with an impaired diffusion capacity [2,3,5,8,11,17,18,2 11. More detailed
0167-8140/93/$06.00 0 1993 Elsevier Scientific Publishers Ireland Ltd. All rights reserved. SSDI
0167-8140(93)01241-G
Image correlation
L.J. Boersma et al. / Radiother. Oncol. 29 (1993)
111
110-116
information on respiratory function can be obtained with planar scintigraphic ventilation (V) and perfusion (Q) scans. Vand Qscans performed 16-407 days after irradiation of the lung showed a marked decrease in Q, followed by a smaller reduction in V [16,21]. The reductions in perfusion and ventilation corresponded with a decrease in overall lung function parameters [6,16,21]. Quantification of regional V and Q with a planar technique, however, is impossible due to overprojection of different regions. Zwijnenburg and coworkers [20,21] used dual-isotope Single Photon Emission Computed Tomography (SPECT) V and Q scans to quantify the lung function of seven patients treated with mantle field irradiation. They found a loss of perfused lung volume of 30%. In conclusion, the available data on radiation-induced lung damage do not give quantitative information on local functional changes, which prevents the determination of dose-effect relations for regional functional changes. These regional dose-effect relations must be known to be able to develop a method to predict the remaining overall pulmonary function of a specific patient from the 3-D dose distribution of that patient [ 151. The aim of this study was, therefore, to develop and validate a method in which the three-dimensional (3-D) dose distribution is combined with 3-D SPECT ventilation and perfusion data, in order to quantify regional changes in relation with the regional radiation dose. This method was validated using data from five patients, treated for Hodgkin’s disease, who were examined prior to radiotherapy and 2-4 months after irradiation, i.e. during the phase of early radiation damage. 2. Methods and patients 2.1. Methods
Single Photon Emission Computed Tomography of the ventilation and perfusion (SPECT \ilQscan) was performed to quantify ventilation and perfusion [9]. A CT scan was used to calculate the three-dimensional (3-D) dose distribution. Both SPECT and CT scans were made prior to radiotherapy and 5 months after the start of the treatment (i.e., 2-4 months after the end of radiotherapy).
per minute (minute ventilation (s)) and the mean volume of air present in the lungs during data acquisition, which must be known for the quantification of the ventilation from the SPECT data. SPECT of the lungs was carried out with the patient lying supine. The arms were raised above the head, in order to avoid disturbances on the count rate caused by the arms. A dualhead gamma camera (Siemens ZLC-75 Rota II, linked to an MDS A3 computer) rotated around the patient. For each camera head, 30 images were obtained at 25 s/image at 6“ intervals. Due to the different energies of the gamma rays from 99mT~(140.5 keV) and ‘lmKr (190 keV), ventilation and perfusion could be determined simultaneously, exploiting the camera’s dual isotope acquisition mode. Sixty-four transaxial slices were reconstructed from the raw SPECT data, using a standard filtered backprojection algorithm. Each slice consisted of 64 x 64 voxels. Standard calibration tests were performed prior to each SPECT examination, showing the pixel size and slice thickness to vary between 5.8 mm and 6.5 mm. The resolution in the transaxial SPECT images was about 1.5 cm. Computed tomography was performed using a Siemens Somatom Plus camera. The slice thickness of the CT cuts was 8 mm, with slices taken every 1 cm, at 1 s/image. During the entire CT scan the patient was in the same position and breathed through the same mouthpiece as during SPECT. Image correlation
The SPECT data were aligned with the corresponding CT data (Fig. 1) in order to apply an attenuation correction of the SPECT data. In addition, the pre- and posttreatment CT data were correlated, in order to align all SPECT data with the 3-D dose distribution, which was connected with the pre-treatment CT data (Fig. 1)
2
Data acquisition
An image of the capillary bed (perfusion) of the lung was obtained after intravenous injection of 185 MBq of ss”‘Tc-labelled (Tlh = 6.0 h) albumin macroaggregates. To obtain a ventilation image, a mixture of *jmKr (T, = 13 s) gas and air was continuously inhaled by the patient through a mouthpiece with low resistance inspiratory and expiratory valves [9]. The expiratory outlet was connected with a pneumotachograph (Jaeger, Wtirzburg, Germany), to measure the total ventilation
Fig. I. Overview of the image-correlation procedures. Both pre- and post-treatment SPECT datasets were aligned with the 3-D dose distribution (pathways la and lb + 2, respectively) which was calculated for the pre-treatment CT data. The SPECT-CT correlations la and I b were performed using external skin markers, the CT-CT correlation 2 was performed using additional anatomical landmarks, if necessary. The correlation between the simulator film and the CT data was carried out using Beams-Eye-View images of the lung contours.
112
L.J. Boersma et al. / Radiother. Oncol. 29 (1993)
Fig. 2. An example of the superimposition of the CT defined lung contour on a SPECT perfusion image after correlation yellow line represents the CT defined lung contour, the red line the perfusion defined lung contour and the ventilation contour blue line.
SPECT-CT correlation was carried out using five external skin markers ( 57Co point sources for SPECT and crossed radio-opaque catheters for CT). A linear transformation matrix was determined by minimizing the root mean square distance (rms distance) between the five corresponding markers. The quality of the match with respect to the position of lungs was visually checked by superimposition of lung contours of the correlated images (Fig. 2). The same method as used for the SPECT-CT correlation, was applied in the CT-CT correlation. The position of the skin markers with respect to the lungs, however, may change during the 5 month interval between the pre- and post-treatment scan, due to possible changes in the outlines of the patient’s body. If comparison of the superimposed lung contours showed a poor match, the anatomical information present in the CT data enabled an improvement to be made in the match by using ad-
110-116
of the datasets. The is represented by the
ditional anatomical landmarks, like the center of a thoracic vertebra, bifurcation of the trachea and the Sterno-clavicular joints. Dose calculation
The 3-D dose distribution was calculated using CT density data from the pre-treatment CT scan, with inhomogeneity corrections based on an equivalent pathlength algorithm (Scandiplan treatment planning system, B.A. Fraass, D.L. McShan, University of Michigan, Ann Arbor, MI). The uncertainty of this algorithm (e.g., caused by errors at the edges of the beams due to the lack of accounting for effects of electron transport) was estimated to be maximally 5%. In our patient group, a problem arose because with mantle field irradiation the position of the patient during treatment (arms angled sideways) was different from the position during CT scanning (arms raised above the
113
L. J. Boersma et al. / Radiother. Oncol. 29 (1993) 110-I I6
head). This change in posture can produce a change in shape and position of the thorax. To find the best position of the radiation field with respect to the lungs in the CT data, the position of the radiation field in the CT dataset was adjusted until the lung contours of the Beams-Eye-View (BEV) images of the lungs showed .an optimal correlation with the lung contours on the simulator film (correlation 3 in Fig. 1). The change in shape of the thorax may also cause dosimetric uncertainties. As a test, we calculated the 3-D dose distribution of a mantle field in a female patient, using CT data with the arms down along the body and CT data with the arms raised above the head. Quantitative analysis of local changes
Prior to quantification of ventilation <\i) and perfusion (o), attenuation correction of the SPECT images was carried out using the method of Chang [ 11,modified for the use of CT densities of the correlated CT data [l!]. Both the perfusion and the ventilation in a voxel i (Qi and &, respectively), were calculated relative to the total pulmonary perfusion and ventilation (Q, and iTE, respectively), within the lung contours defined by the pre-treatment CT scan. In the subsequent analysis, slices in which the diaphragm was visible (i.e., the most caudal 4-6 cm of the lungs) and in which the distance between the lung contours of the pre- and post-treatment CT scans was larger than 1 cm, were excluded to ensure that the analyzed SPECT voxels were located in regions in which anatomical lung tissue was present. In addition, regions with a dose gradient larger than 5 Gylcm were marked because a small spatial inaccuracy in the match in these regions may result in large uncertainties in dose. The post-treatment values of the relative ventilation and perfusion per SPECT voxel were expressed as a percentage of the pre-treatment value. The average relative post treatment value per voxel was calculated per dose interval of 4 Gy, to determine the relation between the radiation dose and the change in the distribution of the ventilation and perfusion. Subsequently, dose-effect relations were calculated in different areas of the lungs in each patient, in order to determine if regional differences were present for the dose-effect relations. 2.2. Patients and treatment details
We have studied the data of five patients (females, aged 17-31 years) irradiated for Hodgkin’s disease. They were treated with mantle field irradiation, using an 8 MV linear accelerator. The total prescribed dose, defined in the mediastinum, was on average 40.4 Gy, given in daily fractions of 1.5-2.0 Gy. In two patients (A and E), the spleen and the para-aortic lymph node region were also irradiated with 40 Gy (fractions of 2 Gy/day). Six courses of chemotherapy prior to the radia-
tion treatment were given to patients B (MOPP/ABV) and D (EBVP). All but one (patient C) had tumor in the mediastinum prior to treatment. Patients were entered in the study after informed consent was acquired. Approval for the study was obtained from the hospital’s ethical committee. 3. Results 3.1. Image correlation The minimized rms-distances between the markers were smaller than 1 cm, both for the CT-CT and the SPECT-CT correlation. Visual comparison of the lung contours, however, showed distances of more than 1.5 cm between the lung contours, in the CT-CT correlations of 3 patients (B, C and D). After adding anatomical markers in these patients, the distances between the lung contours were about 0.5 cm in all CT-CT correlations. Deviations remained larger only in the mediastinal region, particularly in the patients in whom the mediastinum was involved in their Hodgkin’s disease (patients A, B, D and E). In addition, in the basal 4-6 cm of the lung, deviations were seen due to breathing movements of the diaphragm. Visual analysis of the SPECT-CT correlation also showed a good overall match (Fig. 2), except in regions near the diaphragm. In addition, in regions with little perfusion or ventilation (e.g., due to irradiation), the SPECT contours were smaller than the CT contours. These low SPECT values caused differences in the lung contours, mainly in the apices, at the ventral side of the patient and in the irradiated regions. The maximal distances between the SPECT and CT lung contours were found in the mediastinum and lung apices (i.e. in the irradiated areas), and varied from 1 to 7 SPECT voxels (0.6-3.5 cm). 3.2. Dose calculation The maximum uncertainty in position of the mantle fields, expressed as the distances between the lung contours in the simulator film (arms sideways) and the BEV images (arms raised), varied from 0.5 to 0.7 cm both in crania-caudal and lateral directions. For the irradiation fields of the para-aortic lymph nodes and the spleen, these distances varied from 0.5 to 1.2 cm in craniocaudal and lateral directions. These discrepancies were elicited by enlargement of the width of the thorax as the patients raised their arms. The 3-D dose distributions, represented as differential dose volume histograms were different for each patient (Fig.3). In patients A and E doses higher than 52 Gy were found in the regions near the diaphragm, due to the additional para-aortic and spleen fields (Fig. 3). The maximum lung dose in the other patients was found in
114
L.J. Boersma et al. / Radiorher. Oncol. 29 (1993) 110-116
showed that the lung dose in a patient with the arms down was about 3% higher (6% at maximum) than in the situation with the arms raised. In our treatment set-up (with the arms angled sideways) the differences will even be smaller. Together with the uncertainty caused by the calculation algorithm (5%) we had to take into account an uncertainty in dose of maximally 11% in the final analysis. palient E ap
10
20
30
40
50
3.3. Quantitative analysis of local changes
dose(Gy)
30 I
0
10
20
30
40
50
doae(Gy) Fig. 3. Percentage of total number of voxels (total lung volume) per dose interval for each patient (differential dose volume histograms). The solid parts represent the percentage of evaluated voxels, the open parts represent the percentage of voxels, excluded because of a poor match.
the apices of the lungs, being 45-51 Gy. The minimum dose in the lungs was found behind the blocks and was on average 3.6 Gy (6 Gy at maximum). The average dose gradient was larger than 5 Gy/cm in the dose intervals from 12 to 40 Gy (in patient D, from 8 to 36 Gy). In the low- and high-dose regions the average dose gradient was smaller than 5 Gylcm. The results of the test to determine the difference in lung dose caused by the difference between treatment position (arms sideways) and CT position (arms raised)
In nearly all dose intervals some voxels were excluded from the final quantitative analysis (Fig. 3), because they were localized in regions with a poor CT-CT correlation. For most of these excluded voxels the doses were smaller than 12 Gy (near the diaphragm) or higher than 36 Gy (in the apices and, in case of irradiation of the para-aortic lymph nodes and spleen, near the diaphragm). Quantitative analysis showed that the distribution of perfusion (Qi/~) changed with respect to the pretreatment distribution in all patients (Fig. 4). The posttreatment values of perfusion were decreasing with an increasing dose. The curves were different for each patient, but in all patients the dose-effect relations seemed to steepen at about 38 Gy. A decrease of the post-treatment values of the ventilation (&fir) was seen with an increase of dose in patients A-D. The dose-effect relations in these patients also seemed to steepen at 38 Gy. Comparison of the dose-effect relations in different areas of the lungs (left vs. right, ventral vs. caudal, apical vs. basal) did not indicate any influence of the position in the lung on these dose-effect relations. 4. Discussion
We have presented a new method which allows the determination of a dose effect relation for regional radiation-induced functional lung damage in each patient. In this method, the reliability of the image correlations is essential. 4.1. Image correlations
10
20
30
40
50
60
dose(Gy)
Fig. 4. Dose-effect relations for local perfusion (&I&) (solid line) and ventilation (\jE,&) (dashed line), 5 months after irradiation expressed as a percentage of the pre-treatment value, for each patient. The error bars indicate the standard error of the mean (SEM).
The distances between the correlated lung contours ranged from 1 to 7 voxels. These deviations do not necessarily mean a poor match between the anatomical lung structures. Discrepancies in the lung contours are to be expected in parts of the lungs with low SPECT values, as for example in the high-dose regions, because in these regions the lungs may be invisible in the SPECT images. The largest mismatches between the lung contours were indeed seen in the high dose areas (apical and mediastinal). This kind of mismatch will not disturb
L.J. Boersma et al. /Radiother.
Oncol. 29 (1993) 110-116
proper comparison of the local pre- and post-treatment SPECT values in relation with the dose. A geometric mismatch between the anatomical lung structures does cause an uncertainty in the analysis of the local SPECT changes. This kind of mismatch can be due to differences in shape of the lung structures (e.g., caused by breathing movements of the thoracic wall and diaphragm during data acquisition) or by radiationinduced changes in the mediastinum (e.g., shrinkage of mediastinal tumor). Uncertainty in the precise location of the SPECT markers in the images, and a slightly different patient position during the SPECT and CT examinations are other sources of a geometric mismatch. From a detailed study of 10 patients (data not shown), it was concluded that we have to take into account a geometrical matching inaccuracy between the anatomical lung structures of about 0.5 cm (1 SD). A way to improve this accuracy might be to align the images by correlation of the 3-D lung surfaces. We are currently investigating this possibility. Another source of spatial uncertainty in the alignment of the 3-D dose data with the SPECT images is the.difference between treatment and scan position. This difference causes changes in the shape of the internal structures, especially for the PA radiation fields which are given with the patient in the prone position. We did not correct for these changes in shape. The simplest way to resolve this problem would be to irradiate the patients in scan position, but because of the worse reproducibility of this position this is not desirable. 4.2. Interpretation of the quantitative analysis In the intermediate dose regions, the mean dose gradient was larger than 5 Gy/cm. Consequently, small geometric uncertainties could result in large deviations in dose. However, the dose-effect curves (Fig. 4) were rather flat in these dose regions, such that a large difference in dose will give only a small change in effect. The resulting dose-effect relations will therefore hardly be altered by the large dose gradient in these regions. The steepening of the dose-effect relations at approximately 38 Gy coincided with the increase in relative volume at this dose level. It might be suggested that the lung volume in intermediate dose intervals was too small to calculate the changes reliably (Fig. 3). The number of evaluated voxels per dose interval in the intermediate dose ranges (12-36 Gy), however, was still rather large (300 voxels f 110 (1 SD)). The changes of perfusion, as depicted in Fig. 4, represented the changes in the distribution of the blood flow, because all values were normalized to the total value. Assuming that the total pulmonary blood flow would remain constant, the relative increase in the low dose regions indicates a redistribution effect. The
II5
magnitude of this redistribution phenomenon is dependent on the changes in other regions in the lung. Each 3-D dose distribution (Fig. 3) will cause another pattern of local changes, thereby possibly explaining the differences seen between the patients. The same trend of changes as seen for perfusion was also found for the ventilation, but mostly the ventilation changes were somewhat smaller than the perfusion changes. This finding could correspond to a primary change in perfusion, caused by endothelial cell damage [lo], followed by reflex bronchoconstriction in hypoperfused areas [19]. The bleomycin-containing chemotherapy, given in patients B and D prior to irradiation, may provide another explanation for the observed differences. We are currently analyzing more data from different patients to determine the effect of chemotherapy and the effect of the 3-D dose distribution on these dose-effect relations. These data will be used, in combination with data on the overall pulmonary function, to estimate the functional outcome of a specific patient after irradiation, based on the 3-D dose distribution of that specific patient and the regional dose-effect relationships. In these first five patients, examinations were performed only once during the acute phase, to limit the additional examinations for the patients to a minimum. The 5 months follow-up point was chosen as a representative time point for the acute phase; most authors found a rapid progressive decrease in perfusion in the first three months after irradiation [6,16,21], which changed only slightly in the following 9 months. Further study will also include an examination at 18 months after irradiation, during the phase of pulmonary tibrosis. We conclude that the combined use of CT and SPECT information is an effective method to determine the dose-effect relations for regional perfusion and ventilation. Furthermore, this method also makes it possible to measure compensation effects, which is essential to determine the relation between dose, volume and overall functional outcome. 5. Acknowledgements We wish to thank Drs H. Bartelink, A. Begg and B.J. Mijnheer for critical reading of the manuscript, M. Koelman who assisted in processing the data, M. Bakker and other nuclear technologists, who assisted in SPECT data acquisition, N. Verheij, who helped to evaluate the accuracy of the pneumotachograph, and Drs B.A. Fraass and D.L. McShan from the University of Michigan, Ann Arbor, who provided additional software. This work is supported by the Dutch Cancer Society (Grant NKI 90-18).
116
L.J. Boersma et al. / Radiother. Oncol. 29 (1993) 110-116
6. References 1
2
3
4
5
6
Chang, L.-T. A method for attenuation correction in radionuclide computed tomography. IEEE Trans. Nucl. Sci. NS-25: 638-643, 1978. Cionini, L., Pacini, P., De Paola, E., Corradao, A., De Luca Cardillo, C., Mungai, V., Biti, G.P. and Ponticelli, P. Respiratory function tests after mantle irradiation in patients with Hodgkin’s disease. Acta Radiol. 23: 401-409, 1984. do Pica, G.A., Wiley, A.L., Rao, P. and Dickie, H.A. Pulmonary reaction to upper mantle radiation therapy for Hodgkin’s disease. Chest 75: 688-692, 1979. Emami, B., Lyman, J., Brown, A., Coia, L., Goitein, M., Munzenrider, J.E., Shank, B., Solin, L.J. and Wesson, M. Tolerance of normal tissue to therapeutic irradiation. Int. J. Radiat. Oncol., Biol. Phys. 21: 109-122, 1991. Frija, J., Fermi, C., Baud, L., Gisselbrecht, C., Miot, C., Fermand, J.P. and Laval-Jeantet, M. Radiation-induced lung injuries: a survey by computed tomography and pulmonary function tests in 18 cases of Hodgkin’s disease. Eur. J. Radiol. 8: 18-23, 1988. FrBhlich, D., Adler, K. Gottschild, D. and Neumann, G. The value of the pulmonary diffusing capacity in comparison to perfusion scintigraphy of the lung in the radiation pneumonitis. Z. Erkr. Atmungsorgane 173: 170-173, 1989. Gross, N.J. Pulmonary effects of radiation therapy. Ann. Intern. Med. 86: 81-92, 1977. Jones, P.W., Al-Hillawi, A., Waketield, J.M., Johnson, M. and Jelliffe, A.M. Differences in the effect of mediastinal radiotherapy on lung function and the ventilatory response to exercise. Clin. Sci. 67: 389-396, 1984. Klumper, A. and Zwijnenburg, A. Dual isotope (‘lmKr and ss”‘Tc) SPECT in lung function diagnosis. Phys. Med. Biol. 31: 751-761, 1986. Kwock, L., Davenport, W.C., Clark R.L., Zarembra, J., Lingle, B., Chaney, E.L. and Friedman, M. The effects of ionizing radiation on the pulmonary vasculature of intact rats and isolated pulmonary endothelium. Radiat. Res. 11I: 276-291, 1987.
II
12
13
14 I5
16
17
18
19
20
21
Lokich, J.J., Bass, H., Eberly, F.E., Rosenthal, D.S. and Maloney, W.C. The pulmonary effect of mantle irradiation in patients with Hodgkin’s disease. Radiology 108: 397-402, 1973. Mah, K. and Van Dyk, J. Quantitative measurement of changes in human lung density following irradiation. Radiother. Oncol. II: 169-179, 1988. Mah, K., Van Dyk, J., Keane, T. and Poon, P.Y. Acute radiation-induced pulmonary damage: a clinical study on the response to fractionated radiation therapy. Int. J. Radiat. Oncol., Biol. Phys. 13: 179-188, 1987. Mangles, S.H., Jaszczak, R.J., Floyd, C.E., Hahn, L.J., Greer, K.L. and Coleman, R.E. J. Nucl. Med. 28: 1584-1591, 1987. Niemierko, A. and Goitein, M. Modelling of normal tissue response to radiation: the critical volume model. Int. J. Radiat. Oncol., Biol. Phys. 23: 135-145, 1992. Prato, F.S., Kurdyak, R., Sailbil, E.A., Rider, W.D. and Aspin, N. Physiological and radiographic assessment during the development of pulmonary radiation fibrosis. Radiology 122: 389-397, 1977. Shapiro, S.J., Shapiro, SD., Mill, W.B. and Campbell, E.J. Prospective study of long-term pulmonary manifestations of mantle irradiation. Int. J. Radiat. Oncol., Biol. Phys. 19: 707-714, 1990. Smith, L.M., Mendenhall N.P., Cicale, M.J., Block, E.R., Carter, R.L. and Million, R.R. Results of a prospective study evaluating the effects of mantle irradiation on pulmonary function. Int. J. Radiat. Oncol., Biol. Phys. 16: 79-84, 1989. West, J.B. Vascular diseases. In: Pulmonary Pathophysiology, 4th Edn., pp. 108-128. Editors: T.S. Satterfield, L. Napora. Williams and Wilkins, Baltimore, 1992. Zwijnenburg, A. Ventilation-perfusion studies using SPECT. Dissertation, University of Amsterdam, Amsterdam, The Netherlands, 1989. Zwijnenburg, A. Klumper, A. Roos, C.M., van der Schoot, J.B., van Zandwijk, N. and Marcuse, H.R. Lung volume calculations from ‘tmKr SPECT for the quantification of regional ventilation. Chn. Phys. Meas. 9: 147-154, 1988.
A new method to determine dose-effect relations for local lung-function changes using correlated SPECT and CT data L. J. Boersma*a, E.M.F. Damena, R.W. de Boera, S.H. Mulleravb, R.A. ValdCs Olmosb C.A. Hoefnagelb, C.M. RoosC, N. van Zandwijkd, J.V. Lebesque” “Department of Radiotherapy, bDepartment of Nuclear Medicine, dDepartment of Pulmonary Medicine, The Netherlands Cancer Institute (Antoni van Leeuwenhoek Huis), Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands ‘Department of Pulmonary Medicine, Academic Medical Center, Meibergdreef 9, IIOS AZ Amsterdam, The Netherlands
(Received 5 Febuary 1993; revision received 3 May 1993; accepted 17 May 1993)
Abstract Purpose: To determine dose-effect relations for regional lung-function changes after radiotherapy. Methods: Single Photon Emission Computed Tomography (SPECT) was performed to quantify regional ventilation and perfusion. CT scans were used to
calculate the three-dimensional (3-D) dose distribution. Both SPECT and CT scans were performed prior to radiotherapy and 5 months after the start of the treatment. To obtain combined 3-D information on ventilation, perfusion and dose, the SPECT data were correlated with the corresponding CT data. The relative changes in ventilation and perfusion were calculated in each SPECT voxel (voxel size about 6 x 6 x 6 mm) and related to the dose in that voxel. The average relative changes were determined per dose interval of 4 Gy. This procedure was evaluated using the data from five patients treated for Hodgkin’s disease with mantle field irradiation with a prescribed total dose of 40-42 Gy. Results: Dose-effect relations for perfusion were observed in all patients, while in four of the five patients, a dose-effect relation was found for ventilation. The maximal uncertainty of the calculated radiation dose was 11%: a difference between the position of the patient during treatment and during CT scanning caused a maximal dose uncertainty of 6%, while the accuracy of the dose calculation algorithm itself was estimated to be within 5%. Conclusion: The results indicate that the combined use of SPECT and CT information is an effective method for determining dose-effect relations for regional lung function parameters in each individual patient. Key words: Radiation-induced
lung damage; Dose-effect relations; SPECT ventilation/perfusion;
1. Introduction The lung is one of the major dose-limiting organs in treatment of malignant tumors of the thorax. Radiation-induced lung damage occurs in two phases: acute radiation pneumonitis, 1-7 months after treatment, and pulmonary fibrosis, developing from 8 months onwards after radiotherapy 171. The probability and severity of the radiation damage
the radiotherapeutic
depends on radiation dose and on irradiated volume. Emami et al. [4] performed a literature study, from which the total doses with a 5% and 50% complication * Corresponding author.
probability were determined for three different volume categories, with radiation pneumonitis as endpoint (clinical symptoms and characteristic chest X-ray changes). Mah and coworkers [ 12,131 found a doseeffect relation for local damage of lung tissue, as defined by the incidence of CT density changes. To estimate the clinical toxicity, however, it is important to know not only the incidence of radiation pneumonitis or CTdensity changes, but also the functional changes, associated with irradiation of (parts of) the lung. Overall pulmonary function tests, performed within 2 years after irradiation of the lungs, showed a mixed restrictive and obstructive lung disease with an impaired diffusion capacity [2,3,5,8,11,17,18,2 11. More detailed
0167-8140/93/$06.00 0 1993 Elsevier Scientific Publishers Ireland Ltd. All rights reserved. SSDI
0167-8140(93)01241-G
Image correlation
L.J. Boersma et al. / Radiother. Oncol. 29 (1993)
111
110-116
information on respiratory function can be obtained with planar scintigraphic ventilation (V) and perfusion (Q) scans. Vand Qscans performed 16-407 days after irradiation of the lung showed a marked decrease in Q, followed by a smaller reduction in V [16,21]. The reductions in perfusion and ventilation corresponded with a decrease in overall lung function parameters [6,16,21]. Quantification of regional V and Q with a planar technique, however, is impossible due to overprojection of different regions. Zwijnenburg and coworkers [20,21] used dual-isotope Single Photon Emission Computed Tomography (SPECT) V and Q scans to quantify the lung function of seven patients treated with mantle field irradiation. They found a loss of perfused lung volume of 30%. In conclusion, the available data on radiation-induced lung damage do not give quantitative information on local functional changes, which prevents the determination of dose-effect relations for regional functional changes. These regional dose-effect relations must be known to be able to develop a method to predict the remaining overall pulmonary function of a specific patient from the 3-D dose distribution of that patient [ 151. The aim of this study was, therefore, to develop and validate a method in which the three-dimensional (3-D) dose distribution is combined with 3-D SPECT ventilation and perfusion data, in order to quantify regional changes in relation with the regional radiation dose. This method was validated using data from five patients, treated for Hodgkin’s disease, who were examined prior to radiotherapy and 2-4 months after irradiation, i.e. during the phase of early radiation damage. 2. Methods and patients 2.1. Methods
Single Photon Emission Computed Tomography of the ventilation and perfusion (SPECT \ilQscan) was performed to quantify ventilation and perfusion [9]. A CT scan was used to calculate the three-dimensional (3-D) dose distribution. Both SPECT and CT scans were made prior to radiotherapy and 5 months after the start of the treatment (i.e., 2-4 months after the end of radiotherapy).
per minute (minute ventilation (s)) and the mean volume of air present in the lungs during data acquisition, which must be known for the quantification of the ventilation from the SPECT data. SPECT of the lungs was carried out with the patient lying supine. The arms were raised above the head, in order to avoid disturbances on the count rate caused by the arms. A dualhead gamma camera (Siemens ZLC-75 Rota II, linked to an MDS A3 computer) rotated around the patient. For each camera head, 30 images were obtained at 25 s/image at 6“ intervals. Due to the different energies of the gamma rays from 99mT~(140.5 keV) and ‘lmKr (190 keV), ventilation and perfusion could be determined simultaneously, exploiting the camera’s dual isotope acquisition mode. Sixty-four transaxial slices were reconstructed from the raw SPECT data, using a standard filtered backprojection algorithm. Each slice consisted of 64 x 64 voxels. Standard calibration tests were performed prior to each SPECT examination, showing the pixel size and slice thickness to vary between 5.8 mm and 6.5 mm. The resolution in the transaxial SPECT images was about 1.5 cm. Computed tomography was performed using a Siemens Somatom Plus camera. The slice thickness of the CT cuts was 8 mm, with slices taken every 1 cm, at 1 s/image. During the entire CT scan the patient was in the same position and breathed through the same mouthpiece as during SPECT. Image correlation
The SPECT data were aligned with the corresponding CT data (Fig. 1) in order to apply an attenuation correction of the SPECT data. In addition, the pre- and posttreatment CT data were correlated, in order to align all SPECT data with the 3-D dose distribution, which was connected with the pre-treatment CT data (Fig. 1)
2
Data acquisition
An image of the capillary bed (perfusion) of the lung was obtained after intravenous injection of 185 MBq of ss”‘Tc-labelled (Tlh = 6.0 h) albumin macroaggregates. To obtain a ventilation image, a mixture of *jmKr (T, = 13 s) gas and air was continuously inhaled by the patient through a mouthpiece with low resistance inspiratory and expiratory valves [9]. The expiratory outlet was connected with a pneumotachograph (Jaeger, Wtirzburg, Germany), to measure the total ventilation
Fig. I. Overview of the image-correlation procedures. Both pre- and post-treatment SPECT datasets were aligned with the 3-D dose distribution (pathways la and lb + 2, respectively) which was calculated for the pre-treatment CT data. The SPECT-CT correlations la and I b were performed using external skin markers, the CT-CT correlation 2 was performed using additional anatomical landmarks, if necessary. The correlation between the simulator film and the CT data was carried out using Beams-Eye-View images of the lung contours.
112
L.J. Boersma et al. / Radiother. Oncol. 29 (1993)
Fig. 2. An example of the superimposition of the CT defined lung contour on a SPECT perfusion image after correlation yellow line represents the CT defined lung contour, the red line the perfusion defined lung contour and the ventilation contour blue line.
SPECT-CT correlation was carried out using five external skin markers ( 57Co point sources for SPECT and crossed radio-opaque catheters for CT). A linear transformation matrix was determined by minimizing the root mean square distance (rms distance) between the five corresponding markers. The quality of the match with respect to the position of lungs was visually checked by superimposition of lung contours of the correlated images (Fig. 2). The same method as used for the SPECT-CT correlation, was applied in the CT-CT correlation. The position of the skin markers with respect to the lungs, however, may change during the 5 month interval between the pre- and post-treatment scan, due to possible changes in the outlines of the patient’s body. If comparison of the superimposed lung contours showed a poor match, the anatomical information present in the CT data enabled an improvement to be made in the match by using ad-
110-116
of the datasets. The is represented by the
ditional anatomical landmarks, like the center of a thoracic vertebra, bifurcation of the trachea and the Sterno-clavicular joints. Dose calculation
The 3-D dose distribution was calculated using CT density data from the pre-treatment CT scan, with inhomogeneity corrections based on an equivalent pathlength algorithm (Scandiplan treatment planning system, B.A. Fraass, D.L. McShan, University of Michigan, Ann Arbor, MI). The uncertainty of this algorithm (e.g., caused by errors at the edges of the beams due to the lack of accounting for effects of electron transport) was estimated to be maximally 5%. In our patient group, a problem arose because with mantle field irradiation the position of the patient during treatment (arms angled sideways) was different from the position during CT scanning (arms raised above the
113
L. J. Boersma et al. / Radiother. Oncol. 29 (1993) 110-I I6
head). This change in posture can produce a change in shape and position of the thorax. To find the best position of the radiation field with respect to the lungs in the CT data, the position of the radiation field in the CT dataset was adjusted until the lung contours of the Beams-Eye-View (BEV) images of the lungs showed .an optimal correlation with the lung contours on the simulator film (correlation 3 in Fig. 1). The change in shape of the thorax may also cause dosimetric uncertainties. As a test, we calculated the 3-D dose distribution of a mantle field in a female patient, using CT data with the arms down along the body and CT data with the arms raised above the head. Quantitative analysis of local changes
Prior to quantification of ventilation <\i) and perfusion (o), attenuation correction of the SPECT images was carried out using the method of Chang [ 11,modified for the use of CT densities of the correlated CT data [l!]. Both the perfusion and the ventilation in a voxel i (Qi and &, respectively), were calculated relative to the total pulmonary perfusion and ventilation (Q, and iTE, respectively), within the lung contours defined by the pre-treatment CT scan. In the subsequent analysis, slices in which the diaphragm was visible (i.e., the most caudal 4-6 cm of the lungs) and in which the distance between the lung contours of the pre- and post-treatment CT scans was larger than 1 cm, were excluded to ensure that the analyzed SPECT voxels were located in regions in which anatomical lung tissue was present. In addition, regions with a dose gradient larger than 5 Gylcm were marked because a small spatial inaccuracy in the match in these regions may result in large uncertainties in dose. The post-treatment values of the relative ventilation and perfusion per SPECT voxel were expressed as a percentage of the pre-treatment value. The average relative post treatment value per voxel was calculated per dose interval of 4 Gy, to determine the relation between the radiation dose and the change in the distribution of the ventilation and perfusion. Subsequently, dose-effect relations were calculated in different areas of the lungs in each patient, in order to determine if regional differences were present for the dose-effect relations. 2.2. Patients and treatment details
We have studied the data of five patients (females, aged 17-31 years) irradiated for Hodgkin’s disease. They were treated with mantle field irradiation, using an 8 MV linear accelerator. The total prescribed dose, defined in the mediastinum, was on average 40.4 Gy, given in daily fractions of 1.5-2.0 Gy. In two patients (A and E), the spleen and the para-aortic lymph node region were also irradiated with 40 Gy (fractions of 2 Gy/day). Six courses of chemotherapy prior to the radia-
tion treatment were given to patients B (MOPP/ABV) and D (EBVP). All but one (patient C) had tumor in the mediastinum prior to treatment. Patients were entered in the study after informed consent was acquired. Approval for the study was obtained from the hospital’s ethical committee. 3. Results 3.1. Image correlation The minimized rms-distances between the markers were smaller than 1 cm, both for the CT-CT and the SPECT-CT correlation. Visual comparison of the lung contours, however, showed distances of more than 1.5 cm between the lung contours, in the CT-CT correlations of 3 patients (B, C and D). After adding anatomical markers in these patients, the distances between the lung contours were about 0.5 cm in all CT-CT correlations. Deviations remained larger only in the mediastinal region, particularly in the patients in whom the mediastinum was involved in their Hodgkin’s disease (patients A, B, D and E). In addition, in the basal 4-6 cm of the lung, deviations were seen due to breathing movements of the diaphragm. Visual analysis of the SPECT-CT correlation also showed a good overall match (Fig. 2), except in regions near the diaphragm. In addition, in regions with little perfusion or ventilation (e.g., due to irradiation), the SPECT contours were smaller than the CT contours. These low SPECT values caused differences in the lung contours, mainly in the apices, at the ventral side of the patient and in the irradiated regions. The maximal distances between the SPECT and CT lung contours were found in the mediastinum and lung apices (i.e. in the irradiated areas), and varied from 1 to 7 SPECT voxels (0.6-3.5 cm). 3.2. Dose calculation The maximum uncertainty in position of the mantle fields, expressed as the distances between the lung contours in the simulator film (arms sideways) and the BEV images (arms raised), varied from 0.5 to 0.7 cm both in crania-caudal and lateral directions. For the irradiation fields of the para-aortic lymph nodes and the spleen, these distances varied from 0.5 to 1.2 cm in craniocaudal and lateral directions. These discrepancies were elicited by enlargement of the width of the thorax as the patients raised their arms. The 3-D dose distributions, represented as differential dose volume histograms were different for each patient (Fig.3). In patients A and E doses higher than 52 Gy were found in the regions near the diaphragm, due to the additional para-aortic and spleen fields (Fig. 3). The maximum lung dose in the other patients was found in
114
L.J. Boersma et al. / Radiorher. Oncol. 29 (1993) 110-116
showed that the lung dose in a patient with the arms down was about 3% higher (6% at maximum) than in the situation with the arms raised. In our treatment set-up (with the arms angled sideways) the differences will even be smaller. Together with the uncertainty caused by the calculation algorithm (5%) we had to take into account an uncertainty in dose of maximally 11% in the final analysis. palient E ap
10
20
30
40
50
3.3. Quantitative analysis of local changes
dose(Gy)
30 I
0
10
20
30
40
50
doae(Gy) Fig. 3. Percentage of total number of voxels (total lung volume) per dose interval for each patient (differential dose volume histograms). The solid parts represent the percentage of evaluated voxels, the open parts represent the percentage of voxels, excluded because of a poor match.
the apices of the lungs, being 45-51 Gy. The minimum dose in the lungs was found behind the blocks and was on average 3.6 Gy (6 Gy at maximum). The average dose gradient was larger than 5 Gy/cm in the dose intervals from 12 to 40 Gy (in patient D, from 8 to 36 Gy). In the low- and high-dose regions the average dose gradient was smaller than 5 Gylcm. The results of the test to determine the difference in lung dose caused by the difference between treatment position (arms sideways) and CT position (arms raised)
In nearly all dose intervals some voxels were excluded from the final quantitative analysis (Fig. 3), because they were localized in regions with a poor CT-CT correlation. For most of these excluded voxels the doses were smaller than 12 Gy (near the diaphragm) or higher than 36 Gy (in the apices and, in case of irradiation of the para-aortic lymph nodes and spleen, near the diaphragm). Quantitative analysis showed that the distribution of perfusion (Qi/~) changed with respect to the pretreatment distribution in all patients (Fig. 4). The posttreatment values of perfusion were decreasing with an increasing dose. The curves were different for each patient, but in all patients the dose-effect relations seemed to steepen at about 38 Gy. A decrease of the post-treatment values of the ventilation (&fir) was seen with an increase of dose in patients A-D. The dose-effect relations in these patients also seemed to steepen at 38 Gy. Comparison of the dose-effect relations in different areas of the lungs (left vs. right, ventral vs. caudal, apical vs. basal) did not indicate any influence of the position in the lung on these dose-effect relations. 4. Discussion
We have presented a new method which allows the determination of a dose effect relation for regional radiation-induced functional lung damage in each patient. In this method, the reliability of the image correlations is essential. 4.1. Image correlations
10
20
30
40
50
60
dose(Gy)
Fig. 4. Dose-effect relations for local perfusion (&I&) (solid line) and ventilation (\jE,&) (dashed line), 5 months after irradiation expressed as a percentage of the pre-treatment value, for each patient. The error bars indicate the standard error of the mean (SEM).
The distances between the correlated lung contours ranged from 1 to 7 voxels. These deviations do not necessarily mean a poor match between the anatomical lung structures. Discrepancies in the lung contours are to be expected in parts of the lungs with low SPECT values, as for example in the high-dose regions, because in these regions the lungs may be invisible in the SPECT images. The largest mismatches between the lung contours were indeed seen in the high dose areas (apical and mediastinal). This kind of mismatch will not disturb
L.J. Boersma et al. /Radiother.
Oncol. 29 (1993) 110-116
proper comparison of the local pre- and post-treatment SPECT values in relation with the dose. A geometric mismatch between the anatomical lung structures does cause an uncertainty in the analysis of the local SPECT changes. This kind of mismatch can be due to differences in shape of the lung structures (e.g., caused by breathing movements of the thoracic wall and diaphragm during data acquisition) or by radiationinduced changes in the mediastinum (e.g., shrinkage of mediastinal tumor). Uncertainty in the precise location of the SPECT markers in the images, and a slightly different patient position during the SPECT and CT examinations are other sources of a geometric mismatch. From a detailed study of 10 patients (data not shown), it was concluded that we have to take into account a geometrical matching inaccuracy between the anatomical lung structures of about 0.5 cm (1 SD). A way to improve this accuracy might be to align the images by correlation of the 3-D lung surfaces. We are currently investigating this possibility. Another source of spatial uncertainty in the alignment of the 3-D dose data with the SPECT images is the.difference between treatment and scan position. This difference causes changes in the shape of the internal structures, especially for the PA radiation fields which are given with the patient in the prone position. We did not correct for these changes in shape. The simplest way to resolve this problem would be to irradiate the patients in scan position, but because of the worse reproducibility of this position this is not desirable. 4.2. Interpretation of the quantitative analysis In the intermediate dose regions, the mean dose gradient was larger than 5 Gy/cm. Consequently, small geometric uncertainties could result in large deviations in dose. However, the dose-effect curves (Fig. 4) were rather flat in these dose regions, such that a large difference in dose will give only a small change in effect. The resulting dose-effect relations will therefore hardly be altered by the large dose gradient in these regions. The steepening of the dose-effect relations at approximately 38 Gy coincided with the increase in relative volume at this dose level. It might be suggested that the lung volume in intermediate dose intervals was too small to calculate the changes reliably (Fig. 3). The number of evaluated voxels per dose interval in the intermediate dose ranges (12-36 Gy), however, was still rather large (300 voxels f 110 (1 SD)). The changes of perfusion, as depicted in Fig. 4, represented the changes in the distribution of the blood flow, because all values were normalized to the total value. Assuming that the total pulmonary blood flow would remain constant, the relative increase in the low dose regions indicates a redistribution effect. The
II5
magnitude of this redistribution phenomenon is dependent on the changes in other regions in the lung. Each 3-D dose distribution (Fig. 3) will cause another pattern of local changes, thereby possibly explaining the differences seen between the patients. The same trend of changes as seen for perfusion was also found for the ventilation, but mostly the ventilation changes were somewhat smaller than the perfusion changes. This finding could correspond to a primary change in perfusion, caused by endothelial cell damage [lo], followed by reflex bronchoconstriction in hypoperfused areas [19]. The bleomycin-containing chemotherapy, given in patients B and D prior to irradiation, may provide another explanation for the observed differences. We are currently analyzing more data from different patients to determine the effect of chemotherapy and the effect of the 3-D dose distribution on these dose-effect relations. These data will be used, in combination with data on the overall pulmonary function, to estimate the functional outcome of a specific patient after irradiation, based on the 3-D dose distribution of that specific patient and the regional dose-effect relationships. In these first five patients, examinations were performed only once during the acute phase, to limit the additional examinations for the patients to a minimum. The 5 months follow-up point was chosen as a representative time point for the acute phase; most authors found a rapid progressive decrease in perfusion in the first three months after irradiation [6,16,21], which changed only slightly in the following 9 months. Further study will also include an examination at 18 months after irradiation, during the phase of pulmonary tibrosis. We conclude that the combined use of CT and SPECT information is an effective method to determine the dose-effect relations for regional perfusion and ventilation. Furthermore, this method also makes it possible to measure compensation effects, which is essential to determine the relation between dose, volume and overall functional outcome. 5. Acknowledgements We wish to thank Drs H. Bartelink, A. Begg and B.J. Mijnheer for critical reading of the manuscript, M. Koelman who assisted in processing the data, M. Bakker and other nuclear technologists, who assisted in SPECT data acquisition, N. Verheij, who helped to evaluate the accuracy of the pneumotachograph, and Drs B.A. Fraass and D.L. McShan from the University of Michigan, Ann Arbor, who provided additional software. This work is supported by the Dutch Cancer Society (Grant NKI 90-18).
116
L.J. Boersma et al. / Radiother. Oncol. 29 (1993) 110-116
6. References 1
2
3
4
5
6
Chang, L.-T. A method for attenuation correction in radionuclide computed tomography. IEEE Trans. Nucl. Sci. NS-25: 638-643, 1978. Cionini, L., Pacini, P., De Paola, E., Corradao, A., De Luca Cardillo, C., Mungai, V., Biti, G.P. and Ponticelli, P. Respiratory function tests after mantle irradiation in patients with Hodgkin’s disease. Acta Radiol. 23: 401-409, 1984. do Pica, G.A., Wiley, A.L., Rao, P. and Dickie, H.A. Pulmonary reaction to upper mantle radiation therapy for Hodgkin’s disease. Chest 75: 688-692, 1979. Emami, B., Lyman, J., Brown, A., Coia, L., Goitein, M., Munzenrider, J.E., Shank, B., Solin, L.J. and Wesson, M. Tolerance of normal tissue to therapeutic irradiation. Int. J. Radiat. Oncol., Biol. Phys. 21: 109-122, 1991. Frija, J., Fermi, C., Baud, L., Gisselbrecht, C., Miot, C., Fermand, J.P. and Laval-Jeantet, M. Radiation-induced lung injuries: a survey by computed tomography and pulmonary function tests in 18 cases of Hodgkin’s disease. Eur. J. Radiol. 8: 18-23, 1988. FrBhlich, D., Adler, K. Gottschild, D. and Neumann, G. The value of the pulmonary diffusing capacity in comparison to perfusion scintigraphy of the lung in the radiation pneumonitis. Z. Erkr. Atmungsorgane 173: 170-173, 1989. Gross, N.J. Pulmonary effects of radiation therapy. Ann. Intern. Med. 86: 81-92, 1977. Jones, P.W., Al-Hillawi, A., Waketield, J.M., Johnson, M. and Jelliffe, A.M. Differences in the effect of mediastinal radiotherapy on lung function and the ventilatory response to exercise. Clin. Sci. 67: 389-396, 1984. Klumper, A. and Zwijnenburg, A. Dual isotope (‘lmKr and ss”‘Tc) SPECT in lung function diagnosis. Phys. Med. Biol. 31: 751-761, 1986. Kwock, L., Davenport, W.C., Clark R.L., Zarembra, J., Lingle, B., Chaney, E.L. and Friedman, M. The effects of ionizing radiation on the pulmonary vasculature of intact rats and isolated pulmonary endothelium. Radiat. Res. 11I: 276-291, 1987.
II
12
13
14 I5
16
17
18
19
20
21
Lokich, J.J., Bass, H., Eberly, F.E., Rosenthal, D.S. and Maloney, W.C. The pulmonary effect of mantle irradiation in patients with Hodgkin’s disease. Radiology 108: 397-402, 1973. Mah, K. and Van Dyk, J. Quantitative measurement of changes in human lung density following irradiation. Radiother. Oncol. II: 169-179, 1988. Mah, K., Van Dyk, J., Keane, T. and Poon, P.Y. Acute radiation-induced pulmonary damage: a clinical study on the response to fractionated radiation therapy. Int. J. Radiat. Oncol., Biol. Phys. 13: 179-188, 1987. Mangles, S.H., Jaszczak, R.J., Floyd, C.E., Hahn, L.J., Greer, K.L. and Coleman, R.E. J. Nucl. Med. 28: 1584-1591, 1987. Niemierko, A. and Goitein, M. Modelling of normal tissue response to radiation: the critical volume model. Int. J. Radiat. Oncol., Biol. Phys. 23: 135-145, 1992. Prato, F.S., Kurdyak, R., Sailbil, E.A., Rider, W.D. and Aspin, N. Physiological and radiographic assessment during the development of pulmonary radiation fibrosis. Radiology 122: 389-397, 1977. Shapiro, S.J., Shapiro, SD., Mill, W.B. and Campbell, E.J. Prospective study of long-term pulmonary manifestations of mantle irradiation. Int. J. Radiat. Oncol., Biol. Phys. 19: 707-714, 1990. Smith, L.M., Mendenhall N.P., Cicale, M.J., Block, E.R., Carter, R.L. and Million, R.R. Results of a prospective study evaluating the effects of mantle irradiation on pulmonary function. Int. J. Radiat. Oncol., Biol. Phys. 16: 79-84, 1989. West, J.B. Vascular diseases. In: Pulmonary Pathophysiology, 4th Edn., pp. 108-128. Editors: T.S. Satterfield, L. Napora. Williams and Wilkins, Baltimore, 1992. Zwijnenburg, A. Ventilation-perfusion studies using SPECT. Dissertation, University of Amsterdam, Amsterdam, The Netherlands, 1989. Zwijnenburg, A. Klumper, A. Roos, C.M., van der Schoot, J.B., van Zandwijk, N. and Marcuse, H.R. Lung volume calculations from ‘tmKr SPECT for the quantification of regional ventilation. Chn. Phys. Meas. 9: 147-154, 1988.