Purpose/Objective(s)In patients receiving post-prostatectomy radiation therapy (RT), daily image guidance is used to align to the bladder-rectal interface (BRI) which varies depending on bladder and rectal filling. Alignment to pelvic bone is often used for nodal irradiation targets. The impact of alignment prioritized to the BRI on coverage of concurrent nodal targets is unknown. We hypothesized that the nodal clinical target volume (CTVn) would remain within nodal planning target volume (PTV) expansions with alignment to BRI versus bone.Materials/MethodsCone-beam CTs (CBCT) from patients receiving prostate bed RT using a 6 degrees of freedom couch were analyzed using transformation matrices to calculate differences in offsets. The CBCTs with translational offsets of ≥5mm were further evaluated for impact on nodal coverage. The CTVn was contoured according to consensus guidelines and two PTVs generated, using 5 and 7mm symmetric expansions (PTV5=5mm and PTV7=7mm) on the treatment planning CT. Each CBCT of interest was extended superiorly to evaluate the full nodal volumes. Two separate registrations of each CBCT to the treatment planning CT were performed, one with alignment prioritized to the BRI and the other to pelvic bone. The CTVn and PTVs were copied to each CBCT registration (BRI and bone) for analysis (e.g., PTVnBRI is where the planned nodal dose is delivered when the treatment is matched to BRI). The inclusion of the CTVn within each PTV was evaluated. Boolean functions were used to compare the overlap and intersections of the nodal volumes between alignment to BRI versus bone.ResultsA set of 139 CBCTs from 14 patients were initially evaluated, of which 13 CBCTs had maximal translational offsets of ≥5mm. The mean CTVn was 324.3cc, PTV5 was 655.2cc, and PTV7 was 801.4cc. When alignment to the BRI was prioritized, the CTVn was completely included within the PTV7 in 9 of 13 CBCTs. At least 95% of the CTVn was within the PTV5 or PTV7 in all but one CBCT in which 92% of the CTVn was within the PTV5. The volume of the CTVn outside the PTV5 ranged from 2.7-28.4cc (mean 2.6%, range 0.6-8%) and outside the PTV7 from 0-15.4cc (mean 0.7%, range 0-4.3%). The most common location for incomplete inclusion of the CTVn within the PTV was in the most caudal and posterior nodal regions (between the prostate bed and the presacral space), primarily occurring in the event of pitch related mismatch between BRI and bone alignment. When the intersection between PTV(bone) and PTV(BRI) was evaluated, the overlap ranged between 80-98% and did not differ between comparisons for 5 and 7mm expansions.ConclusionThese data from CBCTs with translational offsets of ≥5mm suggest that even with larger disagreement between alignment to BRI versus bone, the vast majority of the CTVn remains encompassed within the PTV, particularly when a 7mm expansion is used. The predominant site of incomplete inclusion raises concern about use of smaller posterior nodal PTV margins. These results are informative for concurrent prostate bed and nodal irradiation.

Radiotherapy treatment (RT) planning for localized prostate cancer is conventionally based upon computed tomography simulation (CT-SIM) which provides an accurate tissue dose estimation. However, the superior soft tissue contrast in magnetic resonance imaging (MRI) offers more accurate delineation of the prostate with some geometric uncertainty. Ideally, both modalities are used simultaneously in practice to ensure the geometric and dosimetric accuracy. In practice, however, only CT-SIM scans are often available, due to accessibility, cost of MR imaging, and speed of CT image acquisition. For improved RT planning, it is useful to have MR images alongside the CT-SIM. Herein, we propose to use a cycle Generative Adversarial Network (CGAN) to produce pelvic MR images from the provided CT-SIM scans in localized prostate cancer patients. For this project, we utilized the CT-SIM and MR images of 34 patients with histopathologically proven prostate cancer treated with RT. After anonymization, each CT-SIM image was windowed with intensity values limited to an interval of (-2000 HU, 2000 HU), and normalized at a patient-level. Similarly, after anonymization, we removed the bias field existing in MR images from the same patient cohort and standardized all intensity distributions. A train-validation split equal to 85%-15% on the patient-level for this study was utilized. As a result, MR and CT slices of 29 patients were used for training the CGAN to learn how to perform a dual cross modality adaptation between CT and MR domains. The C-GAN was trained with a loss function that maintains a spatial registration, between the input CT and the generated MR while transforming. The extracted slices from N = 29 training patients, with 100±28 slices per patient were used to optimize the algorithm (135 epochs, lr = 8e-5). The optimal model was then applied to all CT images in patients from independent validation set. In order to assess the utility of our generated MR images, we asked an expert radiologist to assign a score between 1 to 5 to each slice with higher score indicating the higher quality. Our qualitative investigation showed that we achieved an average score of 4.16 out of 5 in performing CT to MR transformation. MR images generated from CT-SIM images have complete 1:1 spatial correspondence with the CT scans, which facilitates the combined use of MR-CT for a better RT planning. Future work will produce a more generalized model by extending our experiment to our new data-set that incorporates 183 consecutive patients with histopathologically proven prostate cancer. In scenarios in which MRI cannot be obtained due to limited resources, time, or availability, this methodology may hold promise for enhanced target delineation.

The bladder-rectal interface (BRI) position is often dependent on bladder and rectal filling in post-prostatectomy patients. Nodal irradiation is of increasing importance given recent randomized data. It is unknown how alignment to the BRI may impact nodal coverage. We hypothesized that the shifts between alignment to the BRI vs pelvic bone would fall within the standard nodal PTV expansion. CBCTs from patients receiving prostate bed RT using 6 degrees of freedom (6DOF) couch were analyzed. At time of treatment, each CBCT was approved by a prostate specialized radiation oncologist prioritizing alignment to the BRI and prostate bed. Retrospectively each CBCT was re-aligned to pelvic bone. Reference bone points were chosen to represent the superior extent of pelvic nodal target volumes {midline sacral promontory (SP) and anterior aspect of sacroiliac joints at the S1-S2 (LSI and RSI)}. Three cardinal coordinates of each reference point (x, y, and z), were transformed to the machine coordinate system. Transformation matrices were defined by the parameters: lat (x), vert (y), long (z), pitch, roll, and rotation obtained with prospective matching. Using transformation matrices, alignment to the BRI was compared to bone alignment for each reference point by differences between the two sets of coordinates (x"-x'), (y"-y'), (z"-z'). A total of 139 CBCT from 14 patients were evaluated. Differences between coordinates with both alignment techniques were calculated and evaluated across all image sets (see table below). Mean differences between spatial alignment to BRI vs bone were <1.5 mm and >95% of all offsets were ≤5 mm. Approximately 75% of all coordinate differences were ≤1 mm and 90% were ≤3 mm. Less than 4% were >5 mm. Only 2 patients (4 CBCTs) exhibited >7 mm offsets between BRI and bone points. Coordinate differences >3 mm and >5 mm were more common in y (12.5% and 6%) and z (14.6% and 2.9%) than in x dimension (1.4% and 1.4%), respectively. Differences in pitch, roll, and rotation >1° were noted in n = 8, 3, and 1 CBCT sets, respectively. These data suggest that alignment to the prostate bed and BRI did not result in large offsets at the superior aspect of the pelvis. With rare exception, the required shifts should fall within the standard nodal PTV expansion. The impact of bladder and rectal filling on nodal target alignment and coverage is the subject of ongoing investigation.Abstract 4102; TableMean (mm)Std Dev (mm)≤ 1 mm (n)>1-3 mm (n)>3-5 mm (n)>5-7 mm (n)>7 mm (n)Point 1 SPx0.33 (0-8.72)0.951307011y1.40 (0-11.18)1.938140963z0.97 (0-5.92)1.50102191440Point 2 LSIx0.37 (0-8.81)0.981307011y1.39 (0-11.17)1.938042953z0.97 (0-5.93)1.48102161830Point 3 RSIx0.36 (0-8.64)0.941307011y1.40 (0-11.17)1.938141953z1.03 (0-6.20)1.55100171750 Open table in a new tab

Radiation therapy is a cornerstone of treatment for localized prostate cancer (PrCa). As radiation techniques become more precise and larger doses per fraction are utilized, variances in daily set up increasingly become a concern. Guidelines and data on best practice for image guidance, use of implantable devices, and PTV margins is limited. Therefore, we surveyed practicing radiation oncologists to document practice patterns for treatment of PrCa. An IRB approved radiation oncologist-specific practice pattern survey was developed based on current evidence, guidelines, and previously published questionnaires. Survey questions were sent to 1395 ASTRO members who self-identified as treating PrCa between June and October 2018 using SurveyMonkey web-based platform. Conventional fractionation (CFx) was defined as 1.8-2 Gy/fraction and moderate hypofractionation (MHFx) was defined as 2-5 Gy per fraction. Survey responses that had radiation fractionation scheme and >50% of the survey questions completed were included. De-identified participant data was reviewed and Fishers exact test and independent samples T-test were used to compare groups. A total of 159 (11.4%) respondents completed the questionnaire with sufficient information for analysis. 134/159 (84.2%) exclusively used CFx, and 25/159 (15.7%) utilized MHFx. No significant difference for the use of implantable devices was seen in both groups with 113/134 (84.3%) CFx and 19/25 (76%) MHFx (p=0.382). Types of implantable devices used were provided by 107 respondents. Fiducial markers were used by 79/91 (87%) in CFx and 13/16 (81%) with MHFx. Radiofrequency transducers were used by 14/91 (15%) in CFx and 1/16 (6.3%) with MHFx. Hydrogel spacers were used by 41/91 (45%) in CFx and 10/16 (62.5%) with MHFx. For image guidance, the most common protocol was daily cone beam CT (CBCT) by 63% in CFx and 71% in MHFx (p=0.507). Intrafraction motion monitoring systems are used by 16% in CFx and 8% with MHFx. For radiation treatment planning, median posterior, lateral, and superior PTV margins for both CFx and MHFx were 5mm (range 0 – 8), 6mm (range 3 – 15), and 6mm (range 3 – 10) respectively. The variability of data in the literature on effective practice for image guidance, use of implantable devices, and PTV margins for PrCa is reflected in the varied patterns of practice seen in our survey, though there was no difference in recommendations between conventional versus moderately hypofractionated radiation therapy. Consensus guidelines for patterns of practice may be of benefit to radiation oncologist to standardize patterns of care.

Standard imaging is inadequate at identifying recurrent disease post-operatively after prostate cancer biochemical failure. Prostate cancer selectively expresses prostate specific membrane antigen (PSMA). The 18F-DCFBC is a novel radiolabeled PET agent that binds PSMA and provides new information regarding PSMA expression and localization of recurrent disease. We identified 18F-DCFBC PET avid lesions after biochemical failure and compared these to consensus Radiation Therapy Oncology Group (RTOG) volumes for prostate fossa and pelvic lymph nodes (LN). A total of 49 patients enrolled on a prospective institutional review board approved clinical trial. Each presented with biochemical recurrence post-prostatectomy, PSA of ≥0.2ng/ml, and no site of recurrence on standard imaging. All patients underwent whole body 18F-DCFBC PET/CT. RTOG consensus prostate fossa clinical target volume (PF-CTV) contours and pelvic LN volumes (LN-CTV) were drawn by 2 radiation oncologists. PF-CTV limited the superior boarder to 2cm above the pubic symphysis, while the expanded prostate fossa (EPF-CTV) limited the border to 4 cm. The 18F-DCFBC avid lesions were contoured by a radiologist on the corresponding CT while blinded to consensus contours. PET images were fused to CT, and 18F-DCFBC avid areas were compared with consensus contours. Forty-two 18F-DCFBC avid lesions were identified in 26 patients. A total of 40.5% (17/42) of lesions were encompassed within PF-CTV. No lesions were seen within the EPF-CTV. 16.7% (7/42) of lesions were encompassed by LN-CTV. A total of 4.8% (2/42) of recurrent lesions in the pelvis were not within RTOG contours. One lesion was peri-urethral at the level of the penile bulb, while another was identified adjacent to the posterior-lateral rectal wall. A total of 38.1% (16/42) of lesions were classified as distant metastatic disease. Of the 26 patients, 50% (13/26) had recurrent disease within PF-CTV only. 15.4% (4/26) patients had recurrent disease outside of PF-CTV but within LN-CTV. A total of 27.0% (7/26) patients were found to have distant metastatic disease. A total of 7.7% (2/26) patients had local recurrence in areas that were not encompassed by PF-CTV and LN-CTV. Advanced imaging may identify areas of recurrence, inform target volumes, and select appropriate patients for systemic therapy after biochemical failure in prostate cancer. Half of patients had recurrent disease encompassed by consensus radiation volumes for prostate fossa, with only a small number of patients with local disease identified outside this volume. In 15.4% of patients, inclusion of all identified disease would require regional nodal irradiation for adequate coverage. Importantly, 27% of patients had occult metastatic disease and would not have been salvaged.

Pseudo-progression (PsP) has been documented with increasing frequency in glioblastoma (GBM) after adjuvant chemoradiotherapy (CRT). Radiographic features are often not adequate to distinguish PsP from early true progression (eTP); therefore, current response criteria include both clinical and imaging findings in progression assessments. We hypothesize that clinical changes are key to the diagnosis of PsP and delineate the time points in the clinical trajectory that radiographic and clinical changes occur. Sixty-seven patients who received curative intent CRT for GBM at our institute between 2003 and 2015 had pre-treatment and post-treatment imaging suitable for retrospective evaluation using RANO criteria. Patients with signs of progression within the first 12-wks post-radiation (post-RT) were selected. Lesions that subsequently improved or stabilized were denoted as PsP and lesions that continued to progress were considered eTP. Survival was estimated using Kaplan-Meier method, Cox regression was used for multivariate analysis, and Fisher's exact test was used to compare groups. Median follow-up was 17.6 mo. Outcomes revealed a median PFS of 8.0 mo (95% CI: 7.4 – 10.4), and median OS of 20.7 mo (95% CI: 17.3 – 27). Signs of progression developed in 35/67 (52.2%) within 12-wks post-RT. Of these, 20/35 (57.1%) fulfilled criteria for eTP and 15/35 (42.9%) fulfilled criteria for PsP. Time to radiologist-documented radiographic signs of progression were similar between groups at a median of 3.4 weeks for eTP and 3.8 weeks for PsP. At that time, 19 eTP and 15 PsP patients had evaluable MRI. MRI demonstrated similar rates of increased contrast enhancement in 16/19 (84.2%) eTP and 15/15 (100%) PsP. FLAIR was increased, decreased, and stable in 7/15 (46.7%), 4/15 (26.7%), and 4/15 (26.7%) PsP patients, respectively. FLAIR was increased in 14/19 (73.7%) and stable in 5/19 (26.3%) of eTP, with none having decreased FLAIR. A significant difference was seen between PsP and eTP in their need for increased steroid or clinical decline (OR 4.89, 95% CI: 1.003 – 19.27; P = 0.046). 9/20 (45%) ePD and 3/15 (20%) PsP required increased steroid. Importantly, KPS declined in 5/20 (25%) eTP and none of the PsP patients. Unmethylated MGMT was an independent risk factor for survival and predictive of eTP (P = 0.005). As expected, OS was significantly worse in eTP (P = 0.032) with a median OS of 13.2 mo (95% CI: 11.1 – 20.1) versus 23.6 mo (95% CI: 16.5 – 65.3+) for eTP and PsP, respectively. Change in KPS or the need for increased steroids within the first 12-weeks was significantly increased in eTP versus PsP. This was often preceded by clinical neurologic worsening. These findings require prospective validation, and the implementation of standardized assessment of neurologic function would greatly facilitate the widespread use of this metric as an aid in decision making regarding PsP.

The management of glioma has evolved to increasing reliance on molecular characterization over histological findings. We aimed to analyze glioma management and molecular findings in the NCI NIH glioma patient cohort spanning the last 22 years in an effort to elucidate the impact of molecular characterization on response to radiation therapy (RT) as well as additional prognostic features as a platform for ongoing molecular and dosimetric analysis. 102 patients with a diagnosis of grade II and grade III glioma were evaluated between 1995 and 2017, 95 were included in the analysis following thorough chart review. Kaplan-Meier analysis was performed for the purpose of comparison amongst patient and disease variables molecular characterization (IDH mutation, 1p19q deletion), management trends, overall survival (OS) and progression free survival (PFS). 95 patients median age 37 (10-72) with original diagnosis of 21 astrocytoma, 8 Oligoastrocytoma, 20 Olidodendroglioma, 29 Anaplastic Astrocytoma, 3 Anaplastic Oligoastrocytoma, 12 Anaplastic Oligodendroglioma, 2 anaplastic glioneuronal tumor were analyzed. With a median follow-up of 7.3 years, median OS for WHO Grade 2 and Grade 3 cohorts were 8.6 years and 4.8 years (p=.02), with median PFS of 6.9 vs 4.5 years (p=.56). When further stratified by molecular marker subtype, gliomas with 1p/19q co-deletion compared to 1p/19q intact had OS of 7.1 years vs 5.0 (p=.05) and PFS of 5.7 vs 2.6 (p=.22). Gliomas with IDH mutation compared to IDH wild type had OS of 5.5 vs 3.6 years (p=.03) and PFS of 5.3 vs 3.6 years (p=.78). Comparing patients treated with RT versus no RT by histologically based WHO subtype was not associated with a statistically significant difference in OS (RT 7.1 years vs no RT 6.2 years, p=.25) with a trend toward worse PFS in the RT group (RT 5.5 years vs no RT 5.9 years, p=.06). However, when analyzed by molecular subtyping, patients with 1p/19q codeletion and IDH mutated tumors treated with RT had improved OS with 1p/19q co-deleted OS 7.9 years vs 1p/19q intact 4 years, (p=.04). IDH mutated glioma OS was 6.8 years vs IDH wild type 3.85 years (p=.01). PFS analyzed by molecular subtype did not demonstrate a significant difference between groups with 1p/19q co-deleted PFS 6.5 years vs 1p/19q intact 3.7 years, p=.98, and IDH mutated PFS 5.5 years vs IDH wild type 3.5 years, (p=0.62). Previously reported trends were validated in the NCI NIH population. In contrast to histologic WHO grading, molecular characterization was prognostic for OS and predictive for response to RT. The ongoing addition of deep sequencing analysis results to imaging and RT dosimetry characteristics will further define prognostic and predictive factors in the evolving management of this challenging patient population.

Androgen deprivation therapy (ADT) plays an important role in the treatment of men undergoing radiation therapy (RT) for high-risk prostate cancer. Erectile dysfunction (ED) is a common side effect after combined RT and ADT. Both RT and ADT may have cardiovascular effects, thus we hypothesized that penile vascular tissue contracts in response to ADT and RT, and that this change correlates with worsening ED. Penile bulb (PB) volume and corpus cavernosum (CC) width were measured retrospectively on computed tomography (CT) scans obtained pre and post-neoadjuvant ADT (but prior to RT) in a cohort of prostate cancer patients treated with prostate RT+ADT. The same measurements were also obtained from a cohort of male cancer patients treated with RT alone using two CT image sets obtained with similar chronologic spacing. PB volume was measured based on RTOG contouring guidelines. CC width was measured at the position of the anterior border of the inferior pubic ramus. To confirm these findings with a separate image modality and to allow analysis of post-RT changes, we evaluated penile tissue changes with pre-treatment, post-neoadjuvant ADT, and post-RT Magnetic Resonance Imaging (MRI) in prostate cancer patients treated with RT+ADT. Sexual Health Inventory for Men (SHIM) scores were obtained pre-treatment and at least 18 months (range 18-30 months) after discontinuing ADT. Worsening ED was defined as an escalation in SHIM severity classification or initiation of medication for ED. Two-tailed paired t-tests were used to compare differences between cohorts and time points. In the prostate cancer RT+ADT cohort (n=41), both the PB (-3.28 cm3, 95%CI: -4.097 to -2.469; p<0.0001) and the CC (-0.12 cm, 95%CI: -0.1554 to -0.0909; p<0.0001) underwent significant reductions in size in response to neoadjuvant ADT. In the cohort not exposed to ADT (n=20), there was no significant change in PB (+1.38 cm3, 95%CI: -0.9424 to 3.709; p=0.2283) and CC (+0.04 cm, 95%CI: -0.0669 to 0.1510; p=0.4298). Findings from the MRI cohort (n=25) with imaging at three time points confirmed a significant PB reduction after neoadjuvant ADT (-104.4 mm2, 95% CI -137.6 to-71.22, p<0.0001), but not after subsequent treatment with RT (-18.34 mm2. 95% CI -49.81 to 13.13, p=0.2407). In the subset of RT+ADT patients with available SHIM data at late follow up (n=17), there was a nonsignificant trend toward worsening of ED with greater reductions in the percent volume of the PB (p=0.4346) and the percent width of the CC (p=0.112) with neoadjuvant ADT. This retrospective series demonstrates a significant reduction in the volume of erectile tissue with short-term neoadjuvant ADT, and no significant change with the addition of RT. Additional work is needed to determine if these changes persist at later time points or correlate to late ED in larger cohorts.

Multiparametric MRI (mpMRI) has improved detection and risk stratification of men with newly diagnosed prostate cancer. mpMRI has recently been explored in the setting of biochemical recurrence after external beam radiation therapy and High Dose Rate brachytherapy. However, mpMRI is not routinely performed in the setting of biochemical relapse after Low Dose Rate (LDR) brachytherapy because of concern for seed-induced MRI artifact. Herein, we report the first series of post-LDR brachytherapy mpMRI and confirmatory biopsy in the setting of PSA relapse. All patients who received LDR brachytherapy as a component of therapy and were referred to our institution for mpMRI following PSA relapse were analyzed (2011-present). mpMRI consisted of endorectal coil T2 weighted imaging (T2W), dynamic contrast enhanced (DCE) imaging, and diffusion weighted imaging (DWI) sequences with apparent diffusion coefficient (ADC) maps. Lesions were categorized as suspicious by a prostate-dedicated radiologist based on detection in each sequence. Biopsy was performed in patients with no evidence of metastases. Biopsy-proven lesions without unilateral mpMRI findings were considered false negative identifications by mpMRI. Sensitivity (SN) and positive predictive value (PPV) of mpMRI in detecting biopsy-confirmed lesions was determined. 20 patients who developed recurrence post-brachytherapy (Phoenix Criteria: n = 19, rapid PSA rise: n = 1) were included. The mean PSA at the time of mpMRI was 6.42 ng/ml. Pre-implant Gleason score was ≤6 in 12 patients and >6 in 8 patients. The median time from implant to recurrence was 62 months. Prostate biopsy was performed in 17 patients (n = 3 excluded due to metastases). mpMRI detected at least one suspicious prostate lesion in 15/17 patients. A total of 25 lesions were identified (0-4 per patient), 16 of which were confirmed pathologically in 12/17 patients. Pathologic evidence of recurrence was concordant with mpMRI in 12/13 patients. Pathologically confirmed tumors identified by mpMRI were frequently located in the transition zone (TZ, 43.75%) and seminal vesicles (SV, 37.5%). 2/16 of TZ lesions were anterior to the urethra. Two pathologically confirmed SV lesions were not identified by mpMRI. Nine mpMRI defined lesions were benign on biopsy. The overall SN and PPV of detecting pathologically confirmed lesions on any parameter of mpMRI was 70% and 64% respectively. mpMRI detection of local recurrence in the setting of biochemical failure after prostate brachytherapy is feasible, with a high tissue-concordant cancer detection rate. The pattern of local recurrence in the TZ and in the SV post-brachytherapy should be validated in larger series, as it may have implications for treatment planning.

In the past decade, the standard of care for Glioblastoma Multiforme (GBM) has incorporated the use of concurrent and adjuvant temozolomide with radiation therapy (RT) in the post-operative setting (Stupp et al. 2005). Our group has previously published outcome and prognostic features of GBM in patients treated at a single institution, beginning early in the temozolomide era. We are presenting updated outcomes and prognostic features with further time to follow-up and a larger cohort of patients. A total of 114 patients, who received curative intent RT for newly diagnosed GBM at our institute between 1998 and 2015, were reviewed. Treatment consisted of surgical resection followed by adjuvant radiation therapy, with 106 patients receiving concurrent TMZ. For the cohort of 114 patients, the median follow-up time was 19.7 months (range, 0.5-78 months). Median overall survival (OS) and progression free survival (PFS) were 18.6 months, and 7.45 months, respectively. OS for the initial 20 patients treated (most without TMZ) versus the more recent patients, was 13.3 months versus 21.5 months (P = 0.01). 45 patients had MGMT methylation status available, with an OS (36.7 vs 15.3 months; P = 0.003), and PFS (22.2 vs 5.8 months; P = 0.001) benefit seen for methylated versus unmethylated patients. OS and PFS in patients with unknown methylation status, who completed concurrent RT with TMZ on time, were 17.4 months and 7.2 months, respectively. Of patients treated with adjuvant TMZ, 58% were unable to complete at least 6 adjuvant cycles of TMZ chemotherapy, with 25% stopping TMZ due to toxicity and the remainder stopping due to tumor progression. Larger tumor size and periventricular location resulted in decreased median OS (27.7 months vs 15.0 months (P = 0.028), and 23.7 months vs 13.8 months (P = 0.043)). However, with the additional follow-up, there was no benefit in OS or PFS for patients receiving bevacizumab at first recurrence. With increasing follow-up and a larger patient cohort, our clinical data continues to reflect the survival outcomes seen in prospective GBM trials of RT with concurrent TMZ. Tumor size and location continue to be prognostic, but the previously noted benefit of bevacizumab is no longer significant.

Increasing use of re-irradiation (re-RT) for recurrent high grade glioma raises the question of the safely tolerated dose to organs at risk in the field (OAR). Use of manual calculation or treatment planning software to sum plans may result in overestimation of cumulative dose. Deformable registration (DR) has been employed in several tumor sites to better estimate dose and toxicity to OAR when multiple treatment plans are employed. We hypothesize that DR may result in lower cumulative dose to organs at risk in the field and normal tissue complication probability (NTCP) in patients re-irradiated for recurrent high grade glioma as compared to conventional calculation methods. Ten retreatments carried out between 2010 and 2015 were analyzed. Deformable registration (DR) was employed to superimpose the original dose distribution onto the retreatment plan. Maximum and mean dose for optic chiasm, optic nerves, and brain stem were manually extracted from ECLIPSE. NTCP calculations were performed using both sets of data. Wilcoxon matched-pairs signed rank test was employed for statistical analysis. Ten re-treatments were analyzed. Tumor location was frontal (6), parietal (2), temporal (1), occipitoparietal (1). The maximal administered dose to OAR when using DR versus manual summation using treatment planning data differed on average by 6.03 Gy (range -32.13 - 2.56) for the chiasm, 0.5 Gy (range -8.56 – 10.9) and 2.7 Gy (range -7.7 – 0) for right and left optic nerve respectively and by 3 Gy (range -22.4 – 6.69) for the brainstem. A decrease in the calculated dose to OAR was observed in 50-70% of patients with the brainstem and the chiasm exhibiting the largest maximal dose difference with a trend to statistical significance (P = 0.055). A corresponding average decrease in NTCP was noted with 4% for chiasm and right optic nerve, 1% for left optic nerve, and 10% for brainstem. Using mean dose values, the DR dose differed on average by 4 Gy for the chiasm and 1 Gy for right and left optic nerve and brain stem and did not result in a statistically significant change in NTCP. The largest dose differences to OAR using DR were observed in patients whose tumor volumes were overlying the OAR both at first and second treatment. The cumulative maximal dose administered to OAR after re-RT is potentially lower than previously assumed. DR is vulnerable to the definition of the OAR, tumor volume, and location but may provide a more accurate estimation of maximal dose in the re-RT setting. This is especially important as maximal dose continues to form the basis for current dose constraints to OAR in the central nervous system. Patients with the highest doses to OAR stand to benefit most from DR as a means to determine maximal cumulative dose and guide further administration of RT while minimizing NTCP. Further comparison of DR algorithms and increasing numbers of patients in the analysis may help define its role in the setting of re-RT involving the central nervous system.

Conventional MRI is no longer sufficient to accurately identify tumor presence considering the widely documented infiltrative nature of gliomas. Diffusion Weighted Imaging (DWI) has lent further interpretation to available imaging, but data on the optimal combination of DWI sequences to be employed remains elusive. Our intent was to use a convolutional neural network approach to machine learning employing DWI sequences (apparent diffusion coefficient (ADC), relative cerebral blood volume (rCBV)) to identify an imaging biomarker for GBM. Ten histologically documented GBM cases with available detailed operative reports and gross tumor present on DWI prior to the administration of radiation therapy were selected. T1 post-gad images were used to manually delineate tumor which were then coregistered with DWI series. These 10 manually delineated tumors were used to train a convolutional neural network classifier (CNN). In testing, the trained CNN is employed to assign each pixel in the image a probability of belonging to tumor. Receiver Operating Characteristic (ROC) analysis was performed on the probability map to determine optimum thresholds for tumor grading and to obtain the sensitivity, specificity, and positive and negative predictive values for identifying high-grade gliomas. Backtesting of the 10 GBM cases used for machine training achieved almost 100% probability concordance with the T1-gad manually delineated tumors. The trained CNN was then tested on 5 GBM patient datasets with T1-gad, ADC, rCBV images. A sensitivity of 75% and a specificity of 80% with positive and negative predictive values of 79% and 76% respectively were achieved. Optimum threshold for tumor was 0.47. The convolutional neural network approach to DWI analysis may be very useful in identifying a high grade glioma imaging biomarker. Further training with patient data will further improve the accuracy of this approach, enabling its use for recurrence pattern analysis in setting of radiation therapy and systemic treatment, with possible future applicability in tumor grading as well as radiation treatment field design.

Despite mounting evidence for the use of reirradiation (reRT) in recurrent high grade glioma, the safely tolerated dose to organs at risk in the field (OAR) remains unknown. Our intent was to relate dose and volume to normal tissue complication probability (NTCP) with commonly used dose fractionation schemes in reRT. Nine high grade glioma patients (a total of 10 retreatments) retreated between 2010 and 2014 were analyzed. Maximum and equivalent uniform dose (EUD) for optic chiasm, optic nerves and brain stem were extracted from ECLIPSE, and NTCP calculations were performed using historical data and the Lyman model. Chart review and CTCAE v4.0 were employed to define toxicity. Ten retreatments (5 GBM, 5 anaplastic astrocytoma) were analyzed. Mean recurrence GTV and PTV were 27.7 cm3 (0.3-75 cm3) and 123.9 cm3 (35 – 385 cm3), respectively. Initial treatment dose was 55.8-60 Gy (1.8-2 Gy /fraction) with a mean interval between first and second treatment of 3.29 years (1.09 – 5.28). Retreatment dose was 20 – 47.5 Gy (2-3 Gy/fraction). The maximum biological equivalent doses (BED) based on both treatments were 50 Gy3 (10-69), 35 Gy3 (3-74), 34 Gy3 (3-77), and 64 Gy3 (24-82) to chiasm, right optic nerve, left optic nerve and brain stem, respectively. The EUDs to chiasm, right optic nerve, left optic nerve and brainstem were 34 Gy3 (10-47), 22 Gy3 (2-49), 21 Gy3 (2-53), and 28 Gy3 (4-38), respectively. The mean NTCP for chiasm, right optic nerve, left optic nerve and brain stem based on maximal point doses were 26% (0-68%), 17(0-86%), 19%(0-92%), and 53%(0-60%). The mean NTCP for chiasm, right optic nerve, left optic nerve and brainstem based on EUD ranged from 0% to 9%, with the highest probability for brain stem (91%) in one patient with a brain stem glioma. The organs with the most significant probability of NTCP, independent of tumor location, were brainstem followed by optic chiasm. Median time to progression was 2 months (range 0 – 6 months). Median overall survival from retreatment was 6 months (range 1- 15.1 months). None of the patients who received dose to OAR above the published tolerance dose experienced any treatment related grade 3, 4, or 5 toxicities, and all except one patient maintained their pretreatment KPS for at least 2 months following retreatment. The two patients who had the highest NTCP to chiasm, optic nerves and brainstem survived an additional 8 and 15 months from retreatment, respectively. No retreatment related complications were observed. Existing NTCP calculations based on maximal point dose and historical toxicity data may overestimate NTCP in the setting of reRT for recurrent high grade glioma. Since increasing time between treatments and diminished life expectancy likely decrease this probability, it is possible that with careful planning, higher reRT doses can be safely administered in this setting.

Mycosis fungoides (MF) is a rare cutaneous T-cell lymphoma for which topical therapy is the most common initial treatment. Chemotherapy and radiation therapy (RT) are typically reserved for patients who experience disease recurrence. Due to the rarity of this disease, very few prospective studies exist, with RT doses and fractionation largely extrapolated from other disease sites. We assessed salvage RT response rates and toxicities for locally advanced MF patients who have failed all prior local and systematic therapies. We conducted a retrospective review of patient charts and radiation treatment plans from all patients treated at our institution for recurrent cutaneous MF following topical and systemic therapy. Patients with all stages of disease and any number of cutaneous lesions treated between 1996 and 2013 were included. Patients were assessed for local disease control and symptomatic response at the end of treatment. All responses were assessed interpreting the modified severity weighted assessment tool score. One hundred and thirty three lesions from 10 patients treated with external beam RT were included. Lesions were treated with 6MeV (n=31), 9Mev (n=99), 12MeV (n=1) electron beams and 6MV (n=2) photon beams to a mean dose of 12 Gray (Gy) (range 6-36 Gy) in 1.5- 2Gy daily fractions. Prior to RT, 7 patients with 36 lesions progressed on Romidepsin, of which an average of 7 cycles were given (range 1-14 cycles). There were 3 patients with 97 lesions who failed multiple other systematic (mean 2) and local therapies (mean 2). At a median follow-up of 1 month, all lesions demonstrated a complete (CR) or partial response except for one that had a 25% response at the end of treatment but then was lost to follow-up. No difference in response rate was seen according to RT dose administered. The most common toxicity during RT was grade 1 erythema which did not affect patients' ability to complete treatment. There was no acute or late grade ≥2 toxicities observed. Patients treated with external beam RT for recurrent MF following local and systemic therapy uniformly had durable disease response and limited toxicity with a mean dose of 12 Gy in 2 Gy daily fractions. The high rate of disease response and lack of difference seen according to RT dose, support consideration of dose de-escalation even in this challenging cohort.

As dose escalation has been emphasized, photon involved field radiation therapy (IFRT) with chemotherapy has become the standard treatment for locally advanced non-small cell lung cancer (NSCLC). When compared to photon elective nodal irradiation (ENI), photon IFRT has a favorable therapeutic ratio likely secondary to decreased acute toxicity. Given the highly conformal dose distribution achievable with proton radiation therapy, proton ENI could allow for treatment of at-risk nodal regions without increased dose to normal tissues. This study investigates the feasibility of using proton therapy to treat elective nodal stations in patients with locally advanced NSCLC. The plans from 20 patients with locally advanced NSCLC previously treated with photon IFRT to 66.6-72 Gy were evaluated. All of these plans had existing involved field PTVs and normal structure contours. Utilizing an intensity-modulated proton therapy technique, proton IFRT plans were generated to deliver the same total dose as the previously treated photon IFRT plan. For each patient, additional elective nodal CTVs were contoured and expanded to create elective nodal PTVs. Proton ENI plans were generated to deliver 46 Gy to elective nodal PTVs followed by a cone down to original involved field PTVs for a total dose of 66.6-72 Gy. Target coverage and dose to the lung, esophagus, and heart were compared between each of the proton plans and the previously treated photon IFRT plan. Wilcoxon rank-sum test was used to determine significance between the median values for each dosimetric parameter (p≤0.05). Both proton IFRT and proton ENI plans provided significantly improved target volume coverage and decreased doses to the lung, esophagus, and heart compared to the previously treated photon IFRT plans. The median values for dosimetric parameters are as follows:Poster Viewing Abstract 2956; TableDosimetric Parameters for Proton IFRT and Proton ENI Compared to Photon IFRTPhoton IFRTProton IFRTProton ENID95 of involved field PTV95.6%98.7%p=0.00198.7%p=0.002Lung V2029.3%23.6%p=0.00225.0%p=0.03Lung V541.9%31.5%p<0.000138.1%p=0.02Mean lung dose18.2 Gy11.1 Gyp=0.00112.2 Gyp=0.001Esophagus V6039.0%30.5%p=0.00331.4%p=0.001Esophagus V5543.2%34.9%p=0.00634.7%p=0.002Mean esophageal dose38.2 Gy29.7 Gyp=0.000131.4 Gyp=0.005Heart V2533.1%9.2%p=0.00210.2%p=0.002 Open table in a new tab This study demonstrates the feasibility of using proton therapy to treat elective nodal stations in patients with locally advanced NSCLC. Both proton IFRT and proton ENI plans showed superior coverage and normal tissue sparing relative to photon IFRT. Assuming that decreased dose translates to decreased acute toxicity, this study indicates that proton therapy would allow for treatment of at-risk nodal regions while maintaining a favorable therapeutic ratio in comparison to the current standard of photon IFRT.

Intensity modulated radiation therapy (IMRT) is currently used to treat prostate cancer patients with SABR. The purpose of this study was to investigate if predicted normal tissue complication probabilities (NTCP) for the rectum, urethra and bladder differ when comparing IMRT with proton therapy. Anonymized CT data from ten previously treated prostate cancer patients were used to develop radiation treatment plans. IMRT and three different types of proton therapy plans were developed for each patient: opposed lateral plans (2F), opposed laterals with an anterior field (3F), and intensity modulated proton therapy (IMPT). A total of 36.25 Gy was planned to be delivered in 5 fractions. Each plan was normalized to ensure that 95% of the PTV received the prescription dose. The doses to each organ-at-risk were converted to a radiobiologic equivalent dose using accepted α/β ratios for each organ-at-risk (OAR). These equivalent doses were then used to calculate NTCP using a variety of different models including a logistic function model for rectal toxicity (R1), bladder toxicity (B), and urethral toxicity (U1) using parameters described by Takam et al. An additional toxicity analysis of the urethra was performed assuming a uniform dose to the urethra based on the average dose to the prostate (U2). A second analysis of rectal toxicity was performed using a Lyman model for rectal toxicity (R2) with parameters derived by Tucker et al. The NTCPs were analyzed to determine if there was difference in predicted toxicity for each OAR analysis.Poster Viewing Abstract 2512; TableMean NTCPsIMRT2F proton3F protonIMPTR10.27%0.60%0.51%0.23%R20.67%1.43%1.24%0.71%U111.16%12.18%10.09%8.66%U213.32%14.40%11.48%10.52%B0.09%0.11%0.11%0.11% Open table in a new tab IMRT resulted in a statistically significant (SS) lower NTCP using both rectal toxicity models when compared to both 2F (R1 p = .0135; R2 p = .00777) and 3F (R1 p = .0063; R2 p = .0023). There was no SS improvement using IMRT when compared to IMPT on either rectal NTCP analysis. IMPT demonstrated a SS improvement over IMRT for all other NTCP analyses. Proton SABR appears to significantly decrease the predicted rates of urethral toxicity when compared to photon SABR.

Purpose: To develop a GPU‐based interactive multi‐volume visualization program in radiotherapy treatment planning (RTP) which is able to show the spatial relationships between patient anatomical data and radiation dose distribution. Methods: The radiation dose matrix is extracted from commercial RTP systems such as the Eclipse from Varian Oncology Systems (Palo Alto, CA, USA) and the Hi‐Art from the TomoTherapy Incorporated (Madison, WI, USA), and is then co‐registered with the CT volume such that all three volumes of CT data, dose distribution, and segmented radiotherapy structures share the same geometry, resolution, and position, and show no rotation against each other. A GPU‐based multi‐volume ray casting technique is developed by using NVIDIAˈs CUDA framework for simultaneously volume renderings of patient anatomy data and radiation dose distribution. The program is executed on an NVIDIA Tesla C1060 computing processor. Each ray emitted from view point is independently processed by a thread on GPU. The ray transverses both volumes and the visual contributions are mixed for every sample points. Results: The program has been tested on brain tumor patient data and lung tumor patient data. High quality volume rendering of patient anatomy and dose distribution has been generated interactively. The performance of 8 FPS has been achieved for patient data size of 512×512×188, and view window size of 512×512. Conclusions: A multi‐volume visualization program in RTP has been developed on GPU. The program offers visualization through interactive volume rendering of patient anatomical data and radiation dose distribution. The program can be used to improve the understanding of the spatial relationships between patient anatomical data and radiation dose distribution.

Purpose: We performed Monte Carlo simulation using EGSnrc for simulating an X‐RAD 320 Biological Irradiator (Precision X‐Ray Inc., North Branford, CT) which generates orthovoltage x‐rays. Monte‐Carlo simulation was used to determine an appropriate filter to generate the NIST‐traceable beam quality so that a calibration factor necessary for a secondary chamber can be obtained by comparing to a NIST primary standard. Methods: The dimensions and positions of various components including a target within the X‐RAD 320 irradiator were implemented in BEAMnrc. The Monte Carlo simulation was validated by comparing with measurement of x‐ray transmission with a tube potential of 200, 250, and 300 kVp. Using the validated Monte‐Carlo codes, appropriate beam‐hardening filters were investigated to produce the NIST‐traceable beam qualities such as M200, M250, and M300. Results: The simulated transmission of x‐rays with a tube potential of 200, 250, and 300 kVp agreed with the measured ones with a difference of −0.1 ± 2.0%, 1.7 ± 2.1%, and 1.6 ± 2.3%, respectively. The statistical uncertainty in the Monte Carlo simulations was less than 0.5%. The Monte Carlo simulations verified that the NIST‐listed filters for M200 (4.35 mmAl + 1.12 mmCu), M250 (5.25 mmAl + 3.2 mmCu), and M300 (4.25 mmAl + 6.5 mmSn) were adequate for our X‐RAD 320 irradiator. The simulated half‐value layers and homogeneity coefficients using the above filters were 1.64 mmCu and 67.5% for 200 kVp, 3.20 mmCu and 86.1% for 250 kVp, and 5.3 mmCu and 97.1% for 300 kVp, which are in good agreement with the NIST‐traceable beam quality. Conclusions: Monte Carlo simulations of an X‐RAD 320 irradiator were performed to determined beam‐hardening filters for producing the NIST‐traceable beam quality. The NIST‐listed filters were satisfactory for the X‐RAD 320 irradiator to generate the NIST‐traceable beam quality. This research was supported by the Intramural Research Program of the NIH, NCI.