S 184
I MAGE REGISTRATION , SEGMENTATION ALGORITHMS AND DELINEATION
these patients a 0.25 cm CTV-to-PTV margin reduction compared to our clinically prescribed margins of 1 cm is possible, using online CBCT verification. For all other sarcoma subtypes the clinically used PTV margin was too small at the tumour boundaries at normal tissue and air interfaces. 495 poster (Radiotherapy Technologists (RTT) Track) VARIATIONS IN BLADDER VOLUME AND EQUIVALENT UNIFORM DOSE FOR PROSTATE CANCER PATIENTS RECEIVING EXTERNAL RADIOTHERAPY S. Hvaale1 , J. B. Hydal1 , I. Fallmyr1 , K. Jørgensen1 , R. A. H. Nilsskog1 , B. Bø1 , E. Malinen2 1 O SLO U NIVERSITY C OLLEGE, Faculty of Health Sciences, Oslo, Norway 2 O SLO U NIVERSITY H OSPITAL - N ORWEGIAN R ADIUM H OSPITAL, Department of Medical Physics, Division for Cancer Medicine and Radiotherapy, Oslo, Norway
Purpose: To estimate interfraction variations in bladder volume and the resulting variations in equivalent uniform dose (EUD) for prostate cancer patients. Also, to investigate the consequences of drinking practice. Materials: 8 patients with at least 9 CBCT series, obtained during treatment using Elekta’s Synergy XVI, and one planning CT series were included. All patients had fiducial markers implanted in the prostate. The bladder was contoured in each CBCT series and transferred to the planning CT series. Conventional conformal treatment planning was performed using Oncentra Masterplan, providing dose volume histograms for the bladder at each treatment session. This was used to estimate the EUD for each patient and session. Furthermore, prostate rotations were assessed using the fiducials. Results: The interpatient, interfraction mean bladder volume variation was 26 %, while the mean variation in EUD was 3.6 %. Drinking practice had little impact on bladder volume variations and EUD. However, bladder volume variations were significantly associated with prostate rotations. Conclusions: Large bladder volume variations may be expected for prostate cancer patients, irrespective of drinking practice. In any case, variations in EUD were low. Bladder filling may, however, influence prostate rotation. 496 poster (Radiotherapy Technologists (RTT) Track) VOLUMETRIC IMAGING FOR CERVICAL CANCER C. Collen1 , M. Duchateau1 , B. Engels1 , M. De Ridder1 , G. Storme1 1 U NIVERSITAIR Z IEKENHUIS B RUSSEL, Brussels, Belgium
Purpose: Assessment of interfractional organ motion of cervix and uterus by megavoltage computed tomography (MVCT) during intensity-modulated radiotherapy (IMRT). Materials: Ten patients with stage IIB-IVA cervical cancer underwent daily MVCT imaging. Interfractional organ motion was evaluated on 150 pretreatment MVCT images by measuring shifts in their boundaries between the MVCT and planning kV CT scan in the anteroposterior (AP), laterolateral (LL) and superoinferior (SI) directions. Additional intrafractional patient movement was evaluated on 50 post-treatment MVCT images. Results: Measured cervical motion (mean ± SD) was 0.4 mm ± 10.1 mm in anterior, -3.0 mm ± 6.9 mm in posterior direction, -3.5 mm ± 6.9 mm in left and 0.2 mm ± 4.5 mm in right lateral direction, 2.2 mm ± 8.0 mm in superior and 0.5 mm ± 5.0 mm in inferior direction. Compared to the cervix, larger uterine motion was observed. Patient movement during treatment was limited to 1.1 ± 1.25 mm, -0.26 mm ± 1.56 mm, and 0.22 mm ± 2.26 mm in AP, LL and SI direction respectively. Conclusions: MVCT imaging can be used to study patient setup accuracy and internal cervical and uterine motion during IMRT. This data may be used to refine treatment margins.
Image registration, segmentation algorithms and delineation 497 poster (Physics Track) CT VERSUS MRI BASED DELINEATION OF THE CTV BREAST AND CTV BOOST IN BREAST-CONSERVING THERAPY M. Giezen1 , E. Kouwenhoven2 , E. Coerkamp3 , M. Heijenbrok3 , M. Mast1 , A. Petoukhova2 , A. Scholten4 , H. Struikmans1 1 R ADIOTHERAPY C ENTRE W EST, Radiation Oncology, The Hague, Netherlands 2 R ADIOTHERAPY C ENTRE W EST, Medical Physics, The Hague, Netherlands 3 M EDICAL C ENTER H AAGLANDEN, Radiology, The Hague, Netherlands 4 L EIDEN U NIVERSITY M EDICAL C ENTER, Department of Clinical Oncology, Leiden, Netherlands
Purpose: Uncertainty remains whether the glandular breast tissue (GBT) and lumpectomy cavity are correctly delineated when based on CT. MRI might provide a better distinction between fatty degenerated breast tissue and fatty, non breast, tissue. We therefore decided to explore the use of MRI in delineating GBT and lumpectomy cavity for radiation treatment planning. Materials: Five pre- and five postmenopausal patients (cT1-2; N0-1) were scanned by CT and directly afterwards by MRI (no use of intravenous contrast, same patient position) 27±7 days after breast conserving surgery. Two radiation oncologists and 2 radiologists delineated the GBT and lumpectomy cavity on CT as well as on MRI in a "blinded" setting using written delineation instructions. MRI delineations followed a minimum interval of 10 weeks after all CT delineations were completed. GBT and cavity volumes, volume related inter-observer variabilities (IOV’s) and Conformity Indices (CI)1 were quantified for CT and MRI. Cavity visualization scores (CVS) 2 , indicating visibility of a cavity on a scale from 1 to 5, were assessed. Results: The average MRI/CT ratio of GBT volumes is 1.03±0.10. The IOV for delineated GBT volumes is 10%±3% for CT and 13%±4% for MRI. The CI for GBT delineations on CT is 0.81±0.04 and 0.76±0.05 for MRI. The average MRI/CT ratio of cavity volumes is 0.91±0.55. The IOV for lumpectomy cavities is 31%±17% for CT and 78%±43% for MRI. The CI for lumpectomy cavity delineations on CT is 0.53±0.21 and 0.30±0.25 for MRI. Mean difference of scored CVS values between CT and MRI was 0.3±0.8. The IOV for the CVS was 0.5±0.3 for CT and 0.5±0.4 for MRI. CI’s of delineated cavity volumes for CT and MRI show a clear trend with CVS values. Differences in shape and position of GBT and lumpectomy cavity delineations will also be presented. Conclusions: GBT volumes delineated on CT shows no significant difference in mean size, mean IOV and CI in comparison with GBT volumes delineated on MRI. Delineations of lumpectomy cavity volumes show large differences for both CT and MRI. CVS scores are quite similar for CT and MRI with low inter-observer variability for both CT and MRI. 1 Kouwenhoven et al, Phys Med Biol, 2009, in press2 Smitt et al, Radiology, 2001; 219: 203 - 206 498 poster (Physics Track) CT/MRI IMAGE REGISTRATION IN PROSTATE RADIOTHERAPY PLANNING OF PATIENTS WITH UNI- AND BI-LATERAL HIP PROSTHESES J. Brunt1 , A. Haridass2 , Z. Malik2 , C. Eswar Vee2 1 C LATTERBRIDGE C ENTRE FOR O NCOLOGY, Medical Physics, Bebington, Merseyside, United Kingdom 2 C LATTERBRIDGE C ENTRE FOR O NCOLOGY, Bebington, Merseyside, United Kingdom
Purpose: We optimised our MRI protocol for use in (external beam) radiotherapy planning (RTP) of prostate cancer, to minimise errors in image registration with the planning CT. We audited deployment of this for patients with uni- and bi-lateral hip prostheses, to examine the effectiveness of methods of registering CT and MR in the presence of the associated, potentially confounding, localized MR distortion and CT streak artefacts. Materials: Published work recognises factors impeding exact CT/MRI coregistration of pelvic soft tissue for RTP. To minimise inaccuracy, we performed MRI of prostate cancer patients, according to a MRI protocol which utilised a flat table top, knee and ankle supports, and phased array coils, and included for RTP, axial T2W series with 3 mm slices, high spatial resolution in-plane, and sub-millimetre water-fat shift. MR scanning was typically performed less than 90 minutes after CT scanning, itself performed 30 minutes after a microenema. Timing of fluid intake and bladder emptying was specified. Image co-registration of CT/MR using interactive manual and automated (mutual information) methods, was performed on RTP computers. We undertook an audit comprising twenty consecutive patients with unilateral and bilateral hip prostheses, referred for scanning from January 2008 onwards Results: Patient positioning placed the anatomy of interest within 150 mm of the magnet isocentre; in this volume, scanner spatial distortion was found to be less than 2 mm. Of 15 cases where automated registration did not fail completely, detailed registration data were available for 14; comparing automated and interactive registration, mean and maximum differences of translations and rotations were 1.9 mm and 3.1 mm, and 1.4 degrees and 2.3 degrees, respectively. In each of the five cases (25 %) of complete failure, an alternative strategy allowed re-deployment of automated registration with a finally satisfactory result. One strategy for overcoming failures was a "bounding box" automated (BBA) registration (involving manual placement of an active volume excluding the region of the images most affected by artifacts), though there was a case in which this also failed. Provision of an approximate manual interactive registration as a starting point, allowed an automated refinement to succeed in each case (in three cases using BBA). Conclusions: Performing MRI according to our protocol, provides optimized image datasets on which to compare image CT/MR co-registration methods. Even modern automated registration techniques are severely challenged by localized MR distortion and CT streak artefacts arising from hip prostheses, but use of a manually placed "bounding box" allows increased confidence in automated registration. Careful human scrutiny of the results of automated registration remains essential, since failures can occur even when using a "bounding box".