- Email: [email protected]
PD-0353: CT substitute derived from MRI for external beam prostate radiotherapy
ESTRO 33, 2014
S135
work provides a basis for profiling tumours to identify (a combination of) features that allow accurate prediction of treatment response and survival in these patients. This could lead to personalized treatment approaches based on a multi-modal imaging tumour profile. PD-0353 CT substitute derived from MRI for external beam prostate radiotherapy J. Dowling1, P. Pichler2, J. Sun2, D. Rivest-Henault1, S. Ghose1, J. Martin2, C. Wratten2, P. Stanwell3, J. Fripp1, P. Greer2 1 Commonwealth Scientific and Industrial Research Organisation, Australian E-Health Reseach Centre, Brisbane, Australia 2 Calvary Mater Newcastle Hospital, Radiation Oncology, Newcastle, Australia 3 University of Newcastle, School of Health Sciences, Newcastle, Australia Purpose/Objective: The generation of substitute-CT (sCT) from MRI is a requirement for MR-alone radiotherapy planning. The purpose of this study is to derive sCT from an MRI-simulator setup using a multi-atlas framework (also providing automatic organ contouring). Materials and Methods: This study involved 22 patients (mean age 69.4±4.6 years, diagnosed with stage T1-T2 tumors, mean BMI 28.4±3.4) undergoing external beam radiation therapy for localized prostate cancer. Prior to treatment each patient received four MRI sequences: a whole pelvis 1.6mm 3D isotropic T2w (SPACE); and three small field of view scans (encompassing the prostate) consisting of T2w, T2*w and a sequence developed to differentiate fiducial markers from hemorrhage (T1w, TE 6.7 ms, TR 690 ms, flip angle 80°). The MRI scans were acquired from a Siemens Skyra 3T with laser bridge, radiotherapy flat couch top and pelvic coil mounts (CIVCO). Within two hours of MR scanning patients received routine planning CT scans (Toshiba, 2.5mm slices). Contours of the prostate, rectum, bladder, pelvic bones were made by three experienced observers. Each patient CT scan was co-registered to their whole pelvis T2w using symmetric rigid registration followed by structure guided non-rigid registration to maintain bone rigidity while allowing deformation of the bladder and rectum. The sCT conversion involves five steps: First pairwise registration of each whole pelvis MRI in the training database (and associated contours) to the target MRI (affine followed by diffeomorphic demons registration). Secondly the closest atlas images are selected based on similarity (normalized mutual information) to the target MRI. Thirdly contours from the selected atlas cases are combined. The forth step involves applying the same deformation fields from selected atlas scans to the CT-MR coregistered scans. The fifth step involves combining CT-MR results to generate a sCT matching the target MRI. The Dice similarity coefficient (DSC) was used to compare automatic contouring results with manual gold contours. Results: Figure 1 shows the results from a patient T2w MRI scan (A) and the generated sCT (B). The sCT has been rigidly aligned (C) with the patient's planning CT (D). Note internal motion has occurred between scans. DRRs from the MRI-derived sCT (E) and planning CT (F) for this patient are also displayed.
DSC results (mean±sd, n=22) between automatic and manual contouring and inter-observer contouring differences (n=22, 3 observers, planning MRI) are shown in Table 1.
Conclusions: These results are promising and provide necessary tools for MRI-alone treatment planning. Future work will include dose calculations from the sCT scans and shape model refinement of automatic contours. An advantage of this sCT method is that it provides organ contouring and can be applied to standard MRI sequences; however inclusion of additional information is also possible (such as ultra short echo time (UTE) sequences for cortical bone refinement). PD-0354 The 70% of SUVmax threshold on pre radiotherapy PET/CT identifies the site of local recurrence in lung cancer J. Calais1, S. Thureau2, R. Modzelewski3, I. Gardin3, B. Dubray2, P. Vera1 1 Centre Henri Becquerel, Nuclear Medicine, Rouen, France 2 Centre Henri Becquerel, Radiotherapy and Oncology, Rouen, France 3 Centre Henri Becquerel, QuantIF (Litis EA4108), Rouen, France Purpose/Objective: Conventionally-dosed radiotherapy (RT) in nonoperable stage II-III non small cell lung cancer (NSCLC) is associated to high rates of local recurrence. The tolerance of the organs at riks (lung, spinal chord) limits the total RT dose that can be safely administered. The areas with FDG uptake (hotspot) on pre-RT FDG-PET/CT have been reported to identify intra-tumour sites with high risk of relapse after RT [Aerts, Lung Cancer, 2012]. The present study aimed at confirming earlier reports, opening the way to localized dose increase on selected tumour volumes. Materials and Methods: 39 patients with inoperable stage II-III NSCLC, treated with chemo radiation (CTRT) or with RT alone, and who underwent FDG-PET/CT scans at initial staging, before radiotherapy, during radiotherapy (42 Gy), and during systematic follow-up (3 month or 1 year) were extracted from 3 prospective studies (ClinicalTrials.gov Identifier NCT01261585, NCT01261598, RECF0645). All FDG-PET/CT acquisitions were co-registered on the initial scan. The sub volumes in the initial tumour (using 30,40,50,60,70,80 and 90 of SUVmax as thresholds – called B30 to B90) and in the subsequent local recurrence (90% and 40% SUVmax thresholds – called R40, R90), were delineated, pasted on the initial scan and compared for each threshold with overlapping indexes (Dice, Jaccard, overlap fraction, common volume/baseline volume, common volume/recurrent volume)
S135
work provides a basis for profiling tumours to identify (a combination of) features that allow accurate prediction of treatment response and survival in these patients. This could lead to personalized treatment approaches based on a multi-modal imaging tumour profile. PD-0353 CT substitute derived from MRI for external beam prostate radiotherapy J. Dowling1, P. Pichler2, J. Sun2, D. Rivest-Henault1, S. Ghose1, J. Martin2, C. Wratten2, P. Stanwell3, J. Fripp1, P. Greer2 1 Commonwealth Scientific and Industrial Research Organisation, Australian E-Health Reseach Centre, Brisbane, Australia 2 Calvary Mater Newcastle Hospital, Radiation Oncology, Newcastle, Australia 3 University of Newcastle, School of Health Sciences, Newcastle, Australia Purpose/Objective: The generation of substitute-CT (sCT) from MRI is a requirement for MR-alone radiotherapy planning. The purpose of this study is to derive sCT from an MRI-simulator setup using a multi-atlas framework (also providing automatic organ contouring). Materials and Methods: This study involved 22 patients (mean age 69.4±4.6 years, diagnosed with stage T1-T2 tumors, mean BMI 28.4±3.4) undergoing external beam radiation therapy for localized prostate cancer. Prior to treatment each patient received four MRI sequences: a whole pelvis 1.6mm 3D isotropic T2w (SPACE); and three small field of view scans (encompassing the prostate) consisting of T2w, T2*w and a sequence developed to differentiate fiducial markers from hemorrhage (T1w, TE 6.7 ms, TR 690 ms, flip angle 80°). The MRI scans were acquired from a Siemens Skyra 3T with laser bridge, radiotherapy flat couch top and pelvic coil mounts (CIVCO). Within two hours of MR scanning patients received routine planning CT scans (Toshiba, 2.5mm slices). Contours of the prostate, rectum, bladder, pelvic bones were made by three experienced observers. Each patient CT scan was co-registered to their whole pelvis T2w using symmetric rigid registration followed by structure guided non-rigid registration to maintain bone rigidity while allowing deformation of the bladder and rectum. The sCT conversion involves five steps: First pairwise registration of each whole pelvis MRI in the training database (and associated contours) to the target MRI (affine followed by diffeomorphic demons registration). Secondly the closest atlas images are selected based on similarity (normalized mutual information) to the target MRI. Thirdly contours from the selected atlas cases are combined. The forth step involves applying the same deformation fields from selected atlas scans to the CT-MR coregistered scans. The fifth step involves combining CT-MR results to generate a sCT matching the target MRI. The Dice similarity coefficient (DSC) was used to compare automatic contouring results with manual gold contours. Results: Figure 1 shows the results from a patient T2w MRI scan (A) and the generated sCT (B). The sCT has been rigidly aligned (C) with the patient's planning CT (D). Note internal motion has occurred between scans. DRRs from the MRI-derived sCT (E) and planning CT (F) for this patient are also displayed.
DSC results (mean±sd, n=22) between automatic and manual contouring and inter-observer contouring differences (n=22, 3 observers, planning MRI) are shown in Table 1.
Conclusions: These results are promising and provide necessary tools for MRI-alone treatment planning. Future work will include dose calculations from the sCT scans and shape model refinement of automatic contours. An advantage of this sCT method is that it provides organ contouring and can be applied to standard MRI sequences; however inclusion of additional information is also possible (such as ultra short echo time (UTE) sequences for cortical bone refinement). PD-0354 The 70% of SUVmax threshold on pre radiotherapy PET/CT identifies the site of local recurrence in lung cancer J. Calais1, S. Thureau2, R. Modzelewski3, I. Gardin3, B. Dubray2, P. Vera1 1 Centre Henri Becquerel, Nuclear Medicine, Rouen, France 2 Centre Henri Becquerel, Radiotherapy and Oncology, Rouen, France 3 Centre Henri Becquerel, QuantIF (Litis EA4108), Rouen, France Purpose/Objective: Conventionally-dosed radiotherapy (RT) in nonoperable stage II-III non small cell lung cancer (NSCLC) is associated to high rates of local recurrence. The tolerance of the organs at riks (lung, spinal chord) limits the total RT dose that can be safely administered. The areas with FDG uptake (hotspot) on pre-RT FDG-PET/CT have been reported to identify intra-tumour sites with high risk of relapse after RT [Aerts, Lung Cancer, 2012]. The present study aimed at confirming earlier reports, opening the way to localized dose increase on selected tumour volumes. Materials and Methods: 39 patients with inoperable stage II-III NSCLC, treated with chemo radiation (CTRT) or with RT alone, and who underwent FDG-PET/CT scans at initial staging, before radiotherapy, during radiotherapy (42 Gy), and during systematic follow-up (3 month or 1 year) were extracted from 3 prospective studies (ClinicalTrials.gov Identifier NCT01261585, NCT01261598, RECF0645). All FDG-PET/CT acquisitions were co-registered on the initial scan. The sub volumes in the initial tumour (using 30,40,50,60,70,80 and 90 of SUVmax as thresholds – called B30 to B90) and in the subsequent local recurrence (90% and 40% SUVmax thresholds – called R40, R90), were delineated, pasted on the initial scan and compared for each threshold with overlapping indexes (Dice, Jaccard, overlap fraction, common volume/baseline volume, common volume/recurrent volume)