Delineation of brain metastases on CT images for ... - BIR Publications

The British Journal of Radiology, 77 (2004), 39–42 DOI: 10.1259/bjr/68080920

E

2004 The British Institute of Radiology

Delineation of brain metastases on CT images for planning radiosurgery: concerns regarding accuracy 1

K SIDHU, MD, FRCPC, 2P COOPER, MD, FRCPC, 1R RAMANI, PhD, 3M SCHWARTZ, MD, FRCPC, E FRANSSEN, BSc, MSc and 1P DAVEY, MD, FRCPC

1 1

Department of Radiation Oncology, Toronto Sunnybrook Regional Cancer Centre, and Departments of 2Radiology and Neurosurgery, Sunnybrook and Women’s College Health Sciences Centre, University of Toronto, 2075 Bayview Avenue, Toronto, Ontario M4N 3M5 Canada

3

Abstract. Conformal radiotherapy requires confidence that the images used for target delineation accurately reflect the pathological dimensions of the target. Radiosurgery, which is a conformal radiotherapy technique, is often used to treat brain metastases. The images of brain metastases can be affected by the method of image acquisition. A prospective study was undertaken to evaluate the effect of delay on CT images of brain metastases selected for radiosurgical treatment. A median delay from contrast administration of 65 min resulted in an increase in the volume of the metastases in 86% of cases when compared with the volumes of the same metastases determined from CT images acquired immediately following the administration of contrast medium. The magnitude of the increase in volume was sufficient to cause radiosurgery planners to select larger collimator sizes for radiosurgery plans based on the delayed CT images in 92% of cases. No significant intraobserver or interobserver variation was found in the group of radiosurgery planners. Differences in image acquisition may account in part for the differences in local control reported in the radiosurgical treatment of brain metastases.

There has been an exponential growth in the adoption of conformal radiation treatment systems in recent years, which has resulted in the recognition of the problem of target delineation [1]. Radiosurgery is a conformal radiation technique that has been increasingly accepted into clinical practice since the 1980s. Early users of conformal irradiation were aware not only of the arbitrariness of target delineation, but also that treatment planners may disagree among themselves on target delineation even when presented with the same visual information and guidelines for target identification [2–4]. In the case of brain metastases, which are common radiosurgical targets, it has been established in the radiological literature that the dose of contrast agent, as well as the timing of image acquisition following the administration of contrast, can influence the appearance of brain metastases [5–7]. Increased doses of contrast medium, as well as delays in the acquisition of images following contrast medium administration, increases the yield in detecting brain metastases, both in terms of number [8] and size [9]. Accordingly, a prospective study was undertaken to examine the influence of the timing of contrast administration on the appearance of brain metastases selected for radiosurgical treatment, as well as the implications for radiosurgical treatment planning. The study was also designed to provide measures of interobserver and intraobserver variation. Received 11 December 2002 and in revised form 16 July 2003, accepted 20 August 2003. Current address for K Sidhu, Department of Radiation Oncology, Thomas Jefferson Medical College, Thomas Jefferson University Hospital, 111 11th Street, Philadelphia, PA 19107, USA.

The British Journal of Radiology, January 2004

Methods 10 patients with brain metastases referred to the Toronto Sunnybrook Regional Cancer Centre (TSRCC) for radiosurgery consented to participate in the present study, which had been approved by the local research ethics board. Two CT scans were performed on each patient using a GE Spiral CTi scanner (GE Medical Systems, Waukesha, WI) following the administration of iodinated contrast medium. The contrast medium used was a single bolus intravenous injection of 100 cm3 of Omnipaque 300 (Amersham Biosciences, Piscataway, NJ). The first CT scan was acquired immediately after the administration of contrast medium and is described hereafter as the ‘‘immediate scan’’. A second scan was obtained at a median time of 65 min (range 50– 75 min) after the injection of contrast medium. The second scan was labelled the ‘‘delayed scan’’. All CT scans were acquired with an external radiosurgery reference system comprising an Oliver-Bertrand-Tipal (OBT) headframe [10] with CT fiducial plates surgically attached to the patient’s skull. The headframe remained attached for the duration of imaging. The fiducial plates served as markers for a threedimensional reference system. The x, y and z axes represented the anterior to posterior, inferior to superior and right to left direction on the patient, respectively. The Montreal Stereotactic CMITM (McGill University, Montreal, Canada) localization and radiosurgery planning software [11] was used to analyse the images. There were four independent observers who analysed images and planned radiosurgical treatment: a neuroradiologist, a radiation oncologist, a neurosurgeon and a medical physicist. With the exception of the neuroradiologist, all the observers had many years of experience using the Montreal Stereotactic CMITM software. Each observer was asked to contour manually the edge of each tumour image where the edge was defined as the 39

K Sidhu, P Cooper, R Ramani et al

interface between contrast enhancement and surrounding unenhanced brain. The observers were not permitted to alter the windowing or the levelling of the images. Each observer manually contoured each metastasis on three separate occasions at least 1 week apart. The observers were blinded to the timing of the scan, i.e. immediate or delayed. The cross-sectional areas were then converted to a volume by the planning program. A mean difference in the measured volume for the metastasis in the immediate and delayed scan was calculated from the three trials for each of the four observers. The four mean values were then used to calculate an overall mean difference in volume for each of the metastases. The paired sign test was used to evaluate changes in volumes between the immediate and delayed CT scans. The four observers were then asked to prepare a radiosurgery plan for each metastasis. This plan was defined by the choice of the satellite collimator and treatment volume isocentre. There were 11 satellite collimators available ranging from 10 mm to 35 mm in 2.5 mm increments. In order to compare the plans, the prescription was standardized by restricting observers to using the 90% isodose surface which best matched the edge of the image of the metastasis. Each observer generated a plan for each metastasis on three occasions. Finally, the radiosurgery plan that was chosen to actually treat the patient was defined as the reference against which the radiosurgery plans prepared by individual observers were compared. Interobserver and intraobserver variation was determined using Kendall’s coefficient of concordance.

Results A total of 10 patients participated in the study. Four patients had a single brain metastasis and six patients had two metastases, giving a total of 16 metastases. Four of the 10 patients were receiving dexamethasone at the time of the study. Immediate and delayed CT scans were obtained in all 10

patients. In the case of two metastases, contrast was no longer visible on the delayed scan leaving 14 pairs of immediate and delayed images. The observed tumour volumes are summarized in Table 1 and are based on a total of 168 individual datasets. An increase in the volume of the metastases was observed in 12 of 14 instances (p50.006). The mean volume increase in these 12 delayed CT scans was 32% (range 3–70%). In 2 of 14 cases there was a reduction in volume of 1% and 16%, respectively. Radiosurgery plans were generated for the immediate and delayed images on 13 of 14 pairs of images. In one instance, the metastasis exceeded the size of the largest collimator and was excluded from the radiosurgery planning analysis. The results are presented in Table 2 and are based on 156 datasets. The size of the collimators chosen for the delayed CT images increased in 12 of the 13 metastases. In 4 of 13 cases, the observers’ modal choice of collimator was increased by 2 increments (5 mm). The mean distance between the isocentres for the immediate and delayed CT scans was 1.3 mm (range 0.2–4.6 mm). Qualitative differences between the immediate and delayed scans were identified including ring enhancing metastases on immediate scans which became diffusely enhancing on delayed scans and a reduction in the intensity of contrast enhancement in the delayed images (10 of 14 metastases). There were 6 lesions in which a change in the shape of lesion was observed, with no change in shape in the remaining 10 lesions. For one of these lesions (D2), the change in shape necessitated an increase in the size of the collimator, despite an overall decrease in the size of the lesion. Interobserver agreement for collimator size was good for the immediate scans and the delayed scans with a Kendall’s coefficient of concordance of 0.964 and 0.946, respectively. There was no statistically significant difference in each observers repeated trials of radiosurgery planning.

Table 1. Observer measured tumour volumes for CT scans obtained immediately after administration of contrast medium and those obtained after a delay Immediate scan

Delayed scan 3

Lesion

Mean volume (SD) (mm )

A1 A2 B1 B2 C D1 D2 E F1 F2 G1 G2 H1 H2 I J

279 not analysed 290 879 477 1479 1780 1708 1807 2326 1961 3764 5333 not analysed 11358 19787

Mean volume (SD) (mm3)

% change in volume

3D shift in isocentre (mm)

(79)

474

(59)

70

1.4

(87) (114) (15) (32) (33) (33) (21) (23) (161) (234) (138)

325 1134 492 1798 1767 2093 2731 3179 2871 5952 6434

(63) (103) (21) (22) (35) (101) (39) (45) (559) (188) (166)

12 28 3 21 21 22 51 36 46 58 20

1.0 0.7 0.9 1.3 0.4 0.6 0.5 1.5 0.2 1.4 2.4

(344) (894)

13047 16688

(115) (5009)

14 216

4.6 not planned

SD, standard deviation.

40


Brain metastases on CT images Table 2. Collimator sizes chosen by the four observers are listed for each lesion planned using a CT scan obtained immediately after contrast medium administration and planned after a delay in image acquisition. This is compared with the collimator chosen for the actual treatment of the patient Lesion

A1 A2 B1 B2 C D1 D2 E F1 F2 G1 G2 H1 H2 I J

Actual collimator diameter used (mm)

1.25 not analysed 1.25 1.75 1.50 2.00 2.00 1.75 1.75 2.25 2.50 2.25 2.50 not analysed 3.00 not planned

Proposed collimator diameter (mm)

Proposed collimator diameter (mm)

Immediate scan

Delayed scan

Minimum

Maximum

Mode

Minimum

Maximum

Mode

1.00

1.25

1.25

1.00

1.75

1.50

1.00 1.25 1.00 1.50 2.00 1.50 1.75 2.00 1.50 2.00 2.50

1.00 1.75 1.75 2.00 2.25 1.75 2.00 2.25 2.00 2.50 2.75

1.00 1.50 1.25 1.50 2.00 1.50/1.75 1.75/2.00 2.00 1.50/1.75 2.25 2.50

1.00 1.25 1.00 2.00 2.25 2.00 2.00 2.00 1.75 2.50 2.50

1.25 1.75 1.75 2.00 2.75 2.50 2.50 2.50 2.50 2.50 2.75

1.25 1.75 1.25 2.00 2.50 2.00 2.00 2.25 2.25 2.50 2.75

2.75

3.25

3.00

3.25

3.75

3.50

Discussion This study has confirmed previous observations that a delay in the acquisition of CT images following the administration of contrast results in metastases appearing larger than when seen immediately following contrast administration [9]. This study has shown that the magnitude of these changes is sufficient to result in the modification of collimator selection for radiosurgery treatment plans in the majority of cases. The key unanswered question remains: which images best represent the pathological extent of tumour? Although this study was performed with CT imaging, the same principles apply to MR imaging. A recent report of MR based radiosurgery planning [12], describes improved local tumour control when an additional 1 mm margin was added to the gross tumour volume (GTV). 18 month local control rates improved from 54% to 92%. Although the details of the MRI procedure were not given, it may be that the MRI underestimated the true pathological extent of disease, a situation corrected with the use of the 1 mm margin. Local control rates reported for the radiosurgical treatment of brain metastases vary with some series reporting above 90% local control, while other series reporting lower rates [13]. Unfortunately, it remains the case that in much of the radiosurgical literature the methods used for image acquisition are poorly described. The issue of interobserver and intraobserver variation influencing the selection of defined targets [2–4], such as the GTV, is well recognized. In this study, there was excellent interobserver and intraobserver agreement. To a great extent this likely reflects the choice of collimators size available for radiosurgery planning. The collimator size increments (2.5 mm) may have been insufficiently sensitive to reveal subtle disagreements. In the real world of radiosurgery planning, operators have the option of not only selecting a particular collimator size, but of also choosing an isodose volume surface that best covers the target. The British Journal of Radiology, January 2004

In summary, this study has confirmed that the choice of imaging technique can impact on the radiosurgical planning of brain metastases. This may account in part for differences in local control rates reported in published series. Until an optimal imaging protocol is identified, clinical investigators are encouraged to report imaging techniques in detail.

References 1. Leibel SA, Fuks Z, Zelefsky MJ, Wolden SL, Rosenzweig KE, Alektiar KM, et al. Intensity-modulated radiotherapy. Cancer J 2002;8:164–76. 2. Yamamoto M, Nagata Y, Okajima K, Ishigaki T, Murata R, Mizowaki T, et al. Differences in target outline delineation from CT scans of brain tumours using different methods and different observers. Radiother Oncol 1999;50:151–6. 3. Weltens C, Menten J, Feron M, Bellon E, Demaerel P, Maes F, et al. Interoberserver variations in gross tumor volume delineation of brain tumors on computed tomography and impact of magnetic resonance imaging. Radiother Oncol 2001;60:49–59. 4. Aoyama H, Shirato H, Nishioka T, Hashimoto S, Tsuchiya K, Kagei K, et al. Magnetic resonance imaging system for three-dimensional conformal radiotherapy and its impact on gross tumor volume delineation of central nervous system tumors. Int J Radiat Oncol Biol Phys 2001;50:821–7. 5. Runge VM, Kirsch JE, Burke VJ, Price AC, Nelson KL, Thomas GS, et al. High-dose gadoteridol in MR imaging of intracranial neoplasms. JMRI 1992;2:9–18. 6. Yuh WTC, Fisher DJ, Runge VM, Atlas SW, Harms SE, Maravilla KR, et al. Phase III multicenter trial of high-dose gadoteridol in MR evaluation of brain metastases. Am J Neuroradiol 1994;15:1037–51. 7. Bale´riaux D, Colosimo C, Ruscalleda J, Korves M, Schneider G, Bohndorf K, et al. Magnetic resonance imaging of metastatic disease of the brain with gadobenate dimeglumine. Neuroradiology 2002;44:191–203. 8. Shalen PR, Hayman LA, Wallace S, Handel SF. Protocol for delayed contrast enhancement of computed tomography of cerebral neoplasia. Neuroradiology 1981;139:397–402.

41

K Sidhu, P Cooper, R Ramani et al 9. Blatt DR, Friedman WA, Agee OF. Delayed computed tomography contrast enhancement patterns in biopsy proven cases. Neurosurgery 1993;32:560–9. 10. Olivier A, Bertrand G, Peters T. Stereotactic systems and procedures for depth electrode placement: technical aspects. Appl Neurophysiol 1983;46(1–4):37–40. 11. Peters TM, Clark J, Pike B, Drangova M, Olivier A. Stereotactic surgical planning with magnetic resonance imaging, digital substraction angiography and computed tomography. Appl Neurophysiol 1987;50(1–6):33–8.

42

12. Noel G, Valery CA, Proudhom MA, Cornu P, Hasboun D, Boisserie G, et al. Radiosurgery for brain metastasis: comparison of two CTV. Int J Radiat Oncol Biol Phys 2001;51(Suppl. 1):133. 13. Mehta MP. The physical, biologic, and clinical basis of radiosurgery. Curr Probl Cancer 1995;19:265–329.
