- Email: [email protected]
Validation of Diffusion Weighted Magnetic Resonance Imaging for Automatic Tumor Delineation in Laryngeal Cancer
Proceedings of the 53rd Annual ASTRO Meeting Materials/Methods: UT14/UT16A head and neck, and A549 lung tumor xenografts were established subcutaneously in the flank of nu/nu mice. After growth to a minimum diameter of 8mm (n = 5 - 6 per tumor line), tumors underwent PET/CT imaging twice (day 0 and 7) using a FLEX Triumph PET-SPECT-CT system (GE Healthcare/Gamma Medica-Ideas, Waukesha, WI). Mice were anesthetized with 1 - 2% isoflurane and 22.9 MBq (± 10%) of 18F-FDG was injected via tail vein. Animals were dynamically PET imaged for 2 hours followed by a CT scan. Tumor volumes were calculated from CT data. Tumor activity (Q-SUVMAX) from 18F-FDG time-activity curves was used to calculate the Sensitivity Factor, where S-factor = d{ln(Q-SUVMAX)/d{ln(t)} using regression of log-transformed data. The S-factor was correlated with the maximum standardized uptake value (SUVMAX), retention index (RI), traditional kinetic modeling (Ki), specific growth rate (SGR) and tumor doubling time (DT). A p-value \0.05 was considered significant by paired t-tests. Procedures were approved by Beaumont Research Institute IACUC. Results: The S-factor was the only metabolic parameter to significantly change in the UT14/UT16A tumor xenografts in the oneweek interval between individual PET scans (p = 0.009, 0.034). No significant change in the Ki, the RI, or the SUVMAX were evident. There was a strong correlation between the relative changes in tumor volume and those of the S-factor (A549; R = 0.907, UT14; R = 0.828, UT16A; R = 0.689). Interestingly, there was also a strong correlation between the SUVMAX and the Ki in the UT14/UT16A tumors in which the S-factor changed (R = 0.819, 0.701), which was not seen in the A549 tumors that showed no S-factor change. Conclusions: Changes in the S-factor appear to precede changes in the SUVMAX, Ki, and RI, suggesting that the S-factor may be a more informative parameter than the preceding metrics when comparing multiple time points. Additionally, the correlation seen for changes in tumor volume and S-factor suggests that this parameter may be a useful surrogate for tumor growth rate and/or metabolism. This information could be used to incorporate a biological component into radiation treatment planning. Further evaluation of the S-factor is currently underway. Author Disclosure: J.L. Chunta: None. C.O. Wong: None. J. Torma: None. A.A. Martinez: None. B. Marples: None. G.D. Wilson: None.
3093
Performance of a Novel Region Growing Technique for Nuclear Medicine Image Segmentation
R. McGurk, T. J. Smith, S. K. Das Duke University, Durham, NC Purpose/Objective(s): To investigate the performance improvement of a novel region growing (SRG) and ‘‘segmentation-byconsensus’’ (SBC) technique for delineating lesions in FDG-PET images. Further, the improvement in lesion segmentation using time-of-flight (TOF) PET was investigated. Materials/Methods: The NEMA image quality phantom was filled to give 3:1 and 7:1 contrast ratios between spheres (volume = 0.44 - 27.13 cc) and background. Contrast added to the sphere volumes and a hi-resolution CT allowed segmentation of the true activity distribution. Ten minute list-mode scans were acquired on a GE Discovery 690 PET/CT system. Images were reconstructed representing 1, 2 and 5 minute acquisitions using TOF and non-time-of-flight (NTF) methods in order to investigate performance under different noise levels, contrast and segmentation performance using TOF PET. Four commonly used segmentation methods, 40%max 50%max, k-means, fuzzy C-means, and the SRG, were used to segment the spheres. The SRG technique uses an acceptance criteria based on multiple percent thresholds coupled with a search and sort step. Voxels are added to the region if they met the current criteria, the region mean is updated, and voxels now failing the updated criteria are removed. In this fashion the region mean remains relatively high ensuring that the algorithm does not stop prematurely. Further, using SBC, where a volume is created using voxels segmented by a majority of the techniques was investigated. Absolute volume errors (AVE) and volume overlap (VO) were used as metrics to evaluate each algorithm’s performance. Results: Sphere size was the most influential factor as no one technique performed consistently better under different noise levels, contrast or use of TOF. SRG outperformed other techniques for small spheres in low contrast for almost all situations indicated by a mean AVE of 2.30 cc vs. 7.69 - 64.76 cc and 0.47 cc vs. 9.79 - 76.00 cc, and VO of 0.179 vs. 0.006 - 0.014 and 0.454 vs. 0.014 0.166, for the 0.44 cc and 1.08 cc spheres, respectively. For the larger spheres (2.64 - 27.12 cc), no one technique outperformed another. For the high contrast situation, the SRG technique gave the lowest AVE of 2.29 cc vs. 7.99 - 58.29 cc and VO of 0.153 vs. 0.007 - 0.049 for the 0.44 cc sphere. Again, no one technique outperformed others for larger spheres. The SBC technique gave AVE and OM close to the optimal technique for all scenarios. Conclusions: Our SRG technique provides lower mean volume errors, increased volume overlap and less performance variation between sphere volumes in low contrast situations. The SBC is a promising approach and its performance could be improved by incorporating techniques that have been shown to be more robust than simple thresholding. Author Disclosure: R. McGurk: None. T.J. Smith: None. S.K. Das: None.
3094
Validation of Diffusion Weighted Magnetic Resonance Imaging for Automatic Tumor Delineation in Laryngeal Cancer
J. Caldas Magalhaes, N. Kasperts, N. Kooij, C. Terhaard, C. Raaijmakers, M. Philippens UMC Utrecht, Utrecht, Netherlands Purpose/Objective(s): Diffusion weighted magnetic resonance imaging (DW-MRI) is a novel functional imaging technique. It has the advantage of being time- and cost-effective, and does not deliver any radiation dose to the patient. However, due to severe geometric distortions, the role of DW-MRI for tumor delineation has not been investigated in head and neck radiotherapy. We aim
S733
I. J. Radiation Oncology d Biology d Physics
S734
Volume 81, Number 2, Supplement, 2011
to assess the feasibility of automatic gross tumor volume (GTV) delineations in DW-MRI, in laryngeal cancer, and to validate them with pathology. Materials/Methods: DW-MRI (EPI sequence, b-values = 0/150/800 s/mm2) and T2 weighted MRI scans were obtained from 14 patients before total laryngectomy. The DW-MRI scans were visually compared with the T2w MRI scans, to exclude the cases where deformations in the tumor region were larger than 3.5 mm. The GTV was delineated automatically in the DW-MRI b800 images (GTVDWI), using a threshold based on the large gradient of the signal intensity around the tumor. After laryngectomy, the specimen was processed into 3 mm thick-slices and whole mount hematoxylin-eosin (H&E) sections were obtained. A pathologist delineated the tumor in the H&E sections. The specimen was then reconstructed in 3D, and registered with the preoperative imaging with accuracy within 3.5 mm. The coverage of the GTVPATH with the GTVDWI was calculated, according to the formula: GTVcoverage = Intersection (GTVDWI, GTVPATH)/Volume (GTVPATH). Results: 3 of the 14 DW-MRI scans did not present observable geometric distortions, and were selected for the analysis. For these patients, excellent agreement was observed between the tumor delineated in the DW-MRI images and the pathology. Conclusions: DW-MRI based delineations and the tumor on pathology showed similar volumes, and the tumor coverage was 75%. Due to geometric distortions present on the DW-MRI images, the number of patients that could be included in this analysis was limited. Nevertheless, our results indicate that DW-MRI is a promising technique for GTV delineation in laryngeal cancer, when geometric distortions are corrected. Author Disclosure: J. Caldas Magalhaes: None. N. Kasperts: None. N. Kooij: None. C. Terhaard: None. C. Raaijmakers: None. M. Philippens: None.
3095
Translational Development of EPR Oximetry for Assessment of Tumor Hypoxia
B. B. Williams1, L. A. Jarvis1, E. Y. Chen1, H. Hou1, B. I. Zaki1, D. J. Gladstone2, A. C. Hartford1, H. M. Swartz1 1
Dartmouth Medical School, Hanover, NH, 2Dartmouth Medical School, Lebanon, NH
Purpose/Objective(s): Tumor oxygenation is one of the most important factors that affect the response to therapy. Oxygenation changes with disease progression and with therapy in a complex and unpredictable manner, so that direct measurements are needed to follow it under clinically applicable conditions. We aim to provide quantitative pO2 measurements to better enable physicians to characterize disease status and monitor the effects of therapeutic measures. With this information treatments could then be applied with optimal effectiveness by taking into account the oxygen-dependent aspects of the therapy. Materials/Methods: We have developed an approach based on electron paramagnetic resonance (EPR) that makes it feasible to make measurements in the clinical setting under conditions compatible with clinical routines. These measurements are made using low frequency (1.2 GHz) EPR spectroscopy and surface loop resonators which enable measurements at superficial sites using India ink as an oxygen reporter. This technique can be used to provide direct and repeatable measurements of tissue pO2 with a non-invasive measurement procedure following one-time injection of India ink. Ongoing studies in human subjects include oximetry in tumors during courses of radiation and chemo-therapy, and measurement of subcutaneous pO2 in the feet of healthy volunteers to develop procedures that could be used in the treatment of peripheral vascular disease. In addition, we have developed implantable resonators in order to enable measurements of deeper tissues. Results: Oximetry measurements have been performed in tumors of patients during courses of radiation and chemotherapy. Tumor types include melanoma, basal cell, soft tissue sarcoma, and lymphoma, and measurement sites have ranged from the feet to the scalp. Oximetry measurements of subcutaneous tissue on dorsal and plantar foot surfaces have been made in volunteers, with measurements ongoing for each and the longest set of measurements carried out successfully over the last 5 years. In vivo oximetry measurements with implantable resonators were performed in rodent and pig models at several clinically relevant sites, including the brain, to evaluate their performance characteristics and potential for clinical use. Conclusions: These studies demonstrate the feasibility of EPR oximetry in a clinical setting and the potential for more widespread use in the treatment of these and other oxygen dependent diseases. Ongoing developments include the expansion of the technique to determine pO2 in operative fields after irradiation, in wound healing and in restorative oncological surgery, extension of the technique to measure at greater depths using implantable resonators, and the development of capabilities to make the measurements at the bedside and in the operating room. Author Disclosure: B.B. Williams: None. L.A. Jarvis: None. E.Y. Chen: None. H. Hou: None. B.I. Zaki: None. D.J. Gladstone: None. A.C. Hartford: None. H.M. Swartz: None.
3096
Radiosensitization with the PARP Inhibitor ABT-888 Is Independent of PTEN or TP53 Status in Cultured Murine High-grade Astrocytes
L. A. Rosenberg, J. Shields, C. R. Miller University of North Carolina, Chapel Hill, NC Purpose/Objective(s): The majority of high grade astrocytomas (HGA) are ultimately refractory to multimodality therapy. Radiosensitization with poly (ADP-ribose) polymerase (PARP) inhibitors represents a novel therapeutic approach. PARP inhibitors have demonstrated synthetic lethality in cells deficient in DNA repair, including those lacking functional PTEN. Because PTEN and TP53, a gene intimately involved in DNA damage response, are both frequently altered in HGA, we hypothesized that PARP inhibition with would selectively sensitize PTEN- or TP53-deficient HGA to XRT.
3093
Performance of a Novel Region Growing Technique for Nuclear Medicine Image Segmentation
R. McGurk, T. J. Smith, S. K. Das Duke University, Durham, NC Purpose/Objective(s): To investigate the performance improvement of a novel region growing (SRG) and ‘‘segmentation-byconsensus’’ (SBC) technique for delineating lesions in FDG-PET images. Further, the improvement in lesion segmentation using time-of-flight (TOF) PET was investigated. Materials/Methods: The NEMA image quality phantom was filled to give 3:1 and 7:1 contrast ratios between spheres (volume = 0.44 - 27.13 cc) and background. Contrast added to the sphere volumes and a hi-resolution CT allowed segmentation of the true activity distribution. Ten minute list-mode scans were acquired on a GE Discovery 690 PET/CT system. Images were reconstructed representing 1, 2 and 5 minute acquisitions using TOF and non-time-of-flight (NTF) methods in order to investigate performance under different noise levels, contrast and segmentation performance using TOF PET. Four commonly used segmentation methods, 40%max 50%max, k-means, fuzzy C-means, and the SRG, were used to segment the spheres. The SRG technique uses an acceptance criteria based on multiple percent thresholds coupled with a search and sort step. Voxels are added to the region if they met the current criteria, the region mean is updated, and voxels now failing the updated criteria are removed. In this fashion the region mean remains relatively high ensuring that the algorithm does not stop prematurely. Further, using SBC, where a volume is created using voxels segmented by a majority of the techniques was investigated. Absolute volume errors (AVE) and volume overlap (VO) were used as metrics to evaluate each algorithm’s performance. Results: Sphere size was the most influential factor as no one technique performed consistently better under different noise levels, contrast or use of TOF. SRG outperformed other techniques for small spheres in low contrast for almost all situations indicated by a mean AVE of 2.30 cc vs. 7.69 - 64.76 cc and 0.47 cc vs. 9.79 - 76.00 cc, and VO of 0.179 vs. 0.006 - 0.014 and 0.454 vs. 0.014 0.166, for the 0.44 cc and 1.08 cc spheres, respectively. For the larger spheres (2.64 - 27.12 cc), no one technique outperformed another. For the high contrast situation, the SRG technique gave the lowest AVE of 2.29 cc vs. 7.99 - 58.29 cc and VO of 0.153 vs. 0.007 - 0.049 for the 0.44 cc sphere. Again, no one technique outperformed others for larger spheres. The SBC technique gave AVE and OM close to the optimal technique for all scenarios. Conclusions: Our SRG technique provides lower mean volume errors, increased volume overlap and less performance variation between sphere volumes in low contrast situations. The SBC is a promising approach and its performance could be improved by incorporating techniques that have been shown to be more robust than simple thresholding. Author Disclosure: R. McGurk: None. T.J. Smith: None. S.K. Das: None.
3094
Validation of Diffusion Weighted Magnetic Resonance Imaging for Automatic Tumor Delineation in Laryngeal Cancer
J. Caldas Magalhaes, N. Kasperts, N. Kooij, C. Terhaard, C. Raaijmakers, M. Philippens UMC Utrecht, Utrecht, Netherlands Purpose/Objective(s): Diffusion weighted magnetic resonance imaging (DW-MRI) is a novel functional imaging technique. It has the advantage of being time- and cost-effective, and does not deliver any radiation dose to the patient. However, due to severe geometric distortions, the role of DW-MRI for tumor delineation has not been investigated in head and neck radiotherapy. We aim
S733
I. J. Radiation Oncology d Biology d Physics
S734
Volume 81, Number 2, Supplement, 2011
to assess the feasibility of automatic gross tumor volume (GTV) delineations in DW-MRI, in laryngeal cancer, and to validate them with pathology. Materials/Methods: DW-MRI (EPI sequence, b-values = 0/150/800 s/mm2) and T2 weighted MRI scans were obtained from 14 patients before total laryngectomy. The DW-MRI scans were visually compared with the T2w MRI scans, to exclude the cases where deformations in the tumor region were larger than 3.5 mm. The GTV was delineated automatically in the DW-MRI b800 images (GTVDWI), using a threshold based on the large gradient of the signal intensity around the tumor. After laryngectomy, the specimen was processed into 3 mm thick-slices and whole mount hematoxylin-eosin (H&E) sections were obtained. A pathologist delineated the tumor in the H&E sections. The specimen was then reconstructed in 3D, and registered with the preoperative imaging with accuracy within 3.5 mm. The coverage of the GTVPATH with the GTVDWI was calculated, according to the formula: GTVcoverage = Intersection (GTVDWI, GTVPATH)/Volume (GTVPATH). Results: 3 of the 14 DW-MRI scans did not present observable geometric distortions, and were selected for the analysis. For these patients, excellent agreement was observed between the tumor delineated in the DW-MRI images and the pathology. Conclusions: DW-MRI based delineations and the tumor on pathology showed similar volumes, and the tumor coverage was 75%. Due to geometric distortions present on the DW-MRI images, the number of patients that could be included in this analysis was limited. Nevertheless, our results indicate that DW-MRI is a promising technique for GTV delineation in laryngeal cancer, when geometric distortions are corrected. Author Disclosure: J. Caldas Magalhaes: None. N. Kasperts: None. N. Kooij: None. C. Terhaard: None. C. Raaijmakers: None. M. Philippens: None.
3095
Translational Development of EPR Oximetry for Assessment of Tumor Hypoxia
B. B. Williams1, L. A. Jarvis1, E. Y. Chen1, H. Hou1, B. I. Zaki1, D. J. Gladstone2, A. C. Hartford1, H. M. Swartz1 1
Dartmouth Medical School, Hanover, NH, 2Dartmouth Medical School, Lebanon, NH
Purpose/Objective(s): Tumor oxygenation is one of the most important factors that affect the response to therapy. Oxygenation changes with disease progression and with therapy in a complex and unpredictable manner, so that direct measurements are needed to follow it under clinically applicable conditions. We aim to provide quantitative pO2 measurements to better enable physicians to characterize disease status and monitor the effects of therapeutic measures. With this information treatments could then be applied with optimal effectiveness by taking into account the oxygen-dependent aspects of the therapy. Materials/Methods: We have developed an approach based on electron paramagnetic resonance (EPR) that makes it feasible to make measurements in the clinical setting under conditions compatible with clinical routines. These measurements are made using low frequency (1.2 GHz) EPR spectroscopy and surface loop resonators which enable measurements at superficial sites using India ink as an oxygen reporter. This technique can be used to provide direct and repeatable measurements of tissue pO2 with a non-invasive measurement procedure following one-time injection of India ink. Ongoing studies in human subjects include oximetry in tumors during courses of radiation and chemo-therapy, and measurement of subcutaneous pO2 in the feet of healthy volunteers to develop procedures that could be used in the treatment of peripheral vascular disease. In addition, we have developed implantable resonators in order to enable measurements of deeper tissues. Results: Oximetry measurements have been performed in tumors of patients during courses of radiation and chemotherapy. Tumor types include melanoma, basal cell, soft tissue sarcoma, and lymphoma, and measurement sites have ranged from the feet to the scalp. Oximetry measurements of subcutaneous tissue on dorsal and plantar foot surfaces have been made in volunteers, with measurements ongoing for each and the longest set of measurements carried out successfully over the last 5 years. In vivo oximetry measurements with implantable resonators were performed in rodent and pig models at several clinically relevant sites, including the brain, to evaluate their performance characteristics and potential for clinical use. Conclusions: These studies demonstrate the feasibility of EPR oximetry in a clinical setting and the potential for more widespread use in the treatment of these and other oxygen dependent diseases. Ongoing developments include the expansion of the technique to determine pO2 in operative fields after irradiation, in wound healing and in restorative oncological surgery, extension of the technique to measure at greater depths using implantable resonators, and the development of capabilities to make the measurements at the bedside and in the operating room. Author Disclosure: B.B. Williams: None. L.A. Jarvis: None. E.Y. Chen: None. H. Hou: None. B.I. Zaki: None. D.J. Gladstone: None. A.C. Hartford: None. H.M. Swartz: None.
3096
Radiosensitization with the PARP Inhibitor ABT-888 Is Independent of PTEN or TP53 Status in Cultured Murine High-grade Astrocytes
L. A. Rosenberg, J. Shields, C. R. Miller University of North Carolina, Chapel Hill, NC Purpose/Objective(s): The majority of high grade astrocytomas (HGA) are ultimately refractory to multimodality therapy. Radiosensitization with poly (ADP-ribose) polymerase (PARP) inhibitors represents a novel therapeutic approach. PARP inhibitors have demonstrated synthetic lethality in cells deficient in DNA repair, including those lacking functional PTEN. Because PTEN and TP53, a gene intimately involved in DNA damage response, are both frequently altered in HGA, we hypothesized that PARP inhibition with would selectively sensitize PTEN- or TP53-deficient HGA to XRT.