Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review.

Background/purpose: The use of artificial intelligence (AI) in radiotherapy (RT) is expanding rapidly. However, there exists a notable lack of clinician trust in AI models, underscoring the need for effective uncertainty quantification (UQ) methods. The purpose of this study was to scope existing literature related to UQ in RT, identify areas of improvement, and determine future directions. Methods: We followed the PRISMA-ScR scoping review reporting guidelines. We utilized the population (human cancer patients), concept (utilization of AI UQ), context (radiotherapy applications) framework to structure our search and screening process. We conducted a systematic search spanning seven databases, supplemented by manual curation, up to January 2024. Our search yielded a total of 8980 articles for initial review. Manuscript screening and data extraction was performed in Covidence. Data extraction categories included general study characteristics, RT characteristics, AI characteristics, and UQ characteristics. Results: We identified 56 articles published from 2015-2024. 10 domains of RT applications were represented; most studies evaluated auto-contouring (50%), followed by image-synthesis (13%), and multiple applications simultaneously (11%). 12 disease sites were represented, with head and neck cancer being the most common disease site independent of application space (32%). Imaging data was used in 91% of studies, while only 13% incorporated RT dose information. Most studies focused on failure detection as the main application of UQ (60%), with Monte Carlo dropout being the most commonly implemented UQ method (32%) followed by ensembling (16%). 55% of studies did not share code or datasets. Conclusion: Our review revealed a lack of diversity in UQ for RT applications beyond auto-contouring. Moreover, there was a clear need to study additional UQ methods, such as conformal prediction. Our results may incentivize the development of guidelines for reporting and implementation of UQ in RT.

Long-term Patient-Reported Outcomes in a Population-Based Cohort Following Radiotherapy vs Surgery for Oropharyngeal Cancer.

Importance: Oncologic outcomes are similar for patients with oropharyngeal squamous cell carcinoma (OPSCC) treated with primary surgery or radiotherapy. However, comparative differences in long-term patient-reported outcomes (PROs) between modalities are less well established. Objective: To determine the association between primary surgery or radiotherapy and long-term PROs. Design, Setting, and Participants: This cross-sectional study used the Texas Cancer Registry to identify survivors of OPSCC treated definitively with primary radiotherapy or surgery between January 1, 2006, and December 31, 2016. Patients were surveyed in October 2020 and April 2021. Exposures: Primary radiotherapy and surgery for OPSCC. Main Outcomes and Measures: Patients completed a questionnaire that included demographic and treatment information, the MD Anderson Symptom Inventory-Head and Neck (MDASI-HN) module, the Neck Dissection Impairment Index (NDII), and the Effectiveness of Auditory Rehabilitation (EAR) scale. Multivariable linear regression models were performed to evaluate the association of treatment (surgery vs radiotherapy) with PROs while controlling for additional variables. Results: Questionnaires were mailed to 1600 survivors of OPSCC identified from the Texas Cancer Registry, with 400 responding (25% response rate), of whom 183 (46.2%) were 8 to 15 years from their initial diagnosis. The final analysis included 396 patients (aged ≤57 years, 190 [48.0%]; aged >57 years, 206 [52.0%]; female, 72 [18.2%]; male, 324 [81.8%]). After multivariable adjustment, no significant differences were found between surgery and radiotherapy outcomes as measured by the MDASI-HN (ß, -0.1; 95% CI, -0.7 to 0.6), NDII (ß, -1.7; 95% CI, -6.7 to 3.4), and EAR (ß, -0.9; 95% CI -7.7 to 5.8). In contrast, less education, lower household income, and feeding tube use were associated with significantly worse MDASI-HN, NDII, and EAR scores, while concurrent chemotherapy with radiotherapy was associated with worse MDASI-HN and EAR scores. Conclusions and Relevance: This population-based cohort study found no associations between long-term PROs and primary radiotherapy or surgery for OPSCC. Lower socioeconomic status, feeding tube use, and concurrent chemotherapy were associated with worse long-term PROs. Further efforts should focus on the mechanism, prevention, and rehabilitation of these long-term treatment toxicities. The long-term outcomes of concurrent chemotherapy should be validated and may inform treatment decision making.

Real-time MRI motion estimation through an unsupervised k-space-driven deformable registration network (KS-RegNet).

Purpose. Real-time three-dimensional (3D) magnetic resonance (MR) imaging is challenging because of slow MR signal acquisition, leading to highly under-sampled k-space data. Here, we proposed a deep learning-based, k-space-driven deformable registration network (KS-RegNet) for real-time 3D MR imaging. By incorporating prior information, KS-RegNet performs a deformable image registration between a fully-sampled prior image and on-board images acquired from highly-under-sampled k-space data, to generate high-quality on-board images for real-time motion tracking.Methods. KS-RegNet is an end-to-end, unsupervised network consisting of an input data generation block, a subsequent U-Net core block, and following operations to compute data fidelity and regularization losses. The input data involved a fully-sampled, complex-valued prior image, and the k-space data of an on-board, real-time MR image (MRI). From the k-space data, under-sampled real-time MRI was reconstructed by the data generation block to input into the U-Net core. In addition, to train the U-Net core to learn the under-sampling artifacts, the k-space data of the prior image was intentionally under-sampled using the same readout trajectory as the real-time MRI, and reconstructed to serve an additional input. The U-Net core predicted a deformation vector field that deforms the prior MRI to on-board real-time MRI. To avoid adverse effects of quantifying image similarity on the artifacts-ridden images, the data fidelity loss of deformation was evaluated directly in k-space.Results. Compared with Elastix and other deep learning network architectures, KS-RegNet demonstrated better and more stable performance. The average (±s.d.) DICE coefficients of KS-RegNet on a cardiac dataset for the 5- , 9- , and 13-spoke k-space acquisitions were 0.884 ± 0.025, 0.889 ± 0.024, and 0.894 ± 0.022, respectively; and the corresponding average (±s.d.) center-of-mass errors (COMEs) were 1.21 ± 1.09, 1.29 ± 1.22, and 1.01 ± 0.86 mm, respectively. KS-RegNet also provided the best performance on an abdominal dataset.Conclusion. KS-RegNet allows real-time MRI generation with sub-second latency. It enables potential real-time MR-guided soft tissue tracking, tumor localization, and radiotherapy plan adaptation.

Phase II Trial of Sipuleucel-T and Stereotactic Ablative Body Radiation for Patients with Metastatic Castrate-Resistant Prostate Cancer.

(1) We hypothesized that adding concurrent stereotactic ablative radiotherapy (SAbR) would increase the time to progression in patients with metastatic castrate-resistant prostate cancer (mCRPCA) treated with sipuleucel-T. (2) Patients with a history of prostate cancer (PC), radiographic evidence of metastatic disease, and rising prostate-specific antigen (PSA) > 0.2 ng/dL on castrate testosterone levels were enrolled in this single-arm phase II clinical trial and treated with sipuleucel-T and SAbR. The primary endpoint was time to progression (TTP). Cellular and humoral responses were measured using ELISpot and Luminex multiplex assays, respectively. (3) Twenty patients with mCRPC were enrolled and treated with SAbR to 1−3 sites. Treatment was well tolerated with 51, 8, and 4 treatment-related grade 1, 2, and 3 toxicities, respectively, and no grade 4 or 5 adverse events. At a median follow-up of 15.5 months, the median TTP was 11.2 weeks (95% CI; 6.8−14.0 weeks). Median OS was 76.8 weeks (95% CI; 41.6−130.8 weeks). This regimen induced both humoral and cellular immune responses. Baseline M-MDSC levels were elevated in mCRPC patients compared to healthy donors (p = 0.004) and a decline in M-MDSC was associated with biochemical response (p = 0.044). Responders had lower baseline uric acid levels (p = 0.05). No clear correlation with radiographic response was observed. (4) While the regimen was safe, the PC-antigen-specific immune response induced by SAbR did not yield a synergistic clinical benefit for patients treated with sipuleucel-T compared to the historically reported outcomes.

The Prognostic Significance of p16 Status in Patients With Vulvar Cancer Treated With Vulvectomy and Adjuvant Radiation.

PURPOSE: Vulvar squamous cell carcinoma (VSCC) is a relatively rare malignancy. Human papillomavirus has been implicated as a causative factor for a subset of these patients. The purpose of this study was to evaluate whether p16-positivity (a human papillomavirus surrogate) predicts for better response rates in women who undergo surgery followed by adjuvant radiation therapy (RT). METHODS AND MATERIALS: We retrospectively analyzed data from women with VSCC who were treated with adjuvant RT. p16-Positivity was defined as diffuse strong immunoreactivity within the tumor. Time to event outcomes was performed with Kaplan-Meier and cumulative incidence methodologies. RESULTS: Thirty-nine women were identified. Ten had positive results for p16 (p16+), and 29 had negative results (p16-). The median follow-up was 25.7 months. The median age at diagnosis was 59 years for women with p16+ tumors and 74 years for women with p16- tumors (P = .022). The distribution of stage did not differ by p16 status. The indications for adjuvant RT were close/positive margins in 19 women, positive nodes in 9 women, and both in 11 women. There were 21 recurrences: 15 vulvar, 3 isolated nodal, 2 synchronous vulvar/nodal, and 1 distant metastasis. In-field relapse rates at 3 years were lower in p16+ patients (32.5%) than in p16- patients (59.1%, P = .072). This trend was also observed in progression-free survival (P = .062). A p16+ status and a lower International Federation of Gynecology and Obstetrics stage were associated with fewer in-field relapses and improved progression-free survival in multivariable analyses. The p16 status was not a predictor of overall survival. CONCLUSIONS: p16-Positivity appears to be a prognostic factor for in-field relapse rates in patients with VSCC appropriately treated with adjuvant RT.

Thoracic reirradiation with SBRT for residual/recurrent and new primary NSCLC within or immediately adjacent to a prior high-dose radiation field.

PURPOSE: Local failure following concurrent chemoradiation and in-lobe failures following stereotactic body radiation therapy (SBRT) are common. We evaluated our institutional experience using SBRT as salvage in this setting. METHODS AND MATERIALS: Seventy-two patients were reirradiated with SBRT for residual, locally recurrent, or new primary non-small cell lung cancer within or adjacent to a high-dose external beam radiation therapy or SBRT field. Kaplan-Meier analysis with log-rank test were used to estimate endpoints and differentiate cohorts. RESULTS: Median follow-up was 17.9 months. Patients had residual or recurrent disease (54.2%); 45.8% had new lung primaries. Median reirradiated T size was 2.5 cm (range, 0.8-7.8 cm). Median pre-retreatment maximum standardized uptake value (SUVmax) was 7.15 (range, 1.2-37.6). The most common SBRT reirradiation regimen was 48 Gy in 4 fractions (range, 17-60 Gy in 1-5 fractions). Median progression-free survival was 15.2 months, and median overall survival was 20.8 months. Two-year local failure was 21.6%. Patients with SUVmax at reirradiation <7.0 had a 2-year local control of 93.1% versus 61.1% above the median (P < .001). The 2-year rate of distant metastases was 10.4% versus 54.1% in patients treated for a new primary versus residual or recurrent disease (P < .001). Median progression-free survival was 31.9 months versus 8.4 months, respectively (P = .037). Median survival of patients treated for new primary was 25.2 months versus 16.2 months with residual or recurrent disease (P = .049), and median survival for patients with reirradiation SUVmax below the median was 42.0 months versus 9.8 months above the median (P < .001). Acute any-grade toxicity was seen in 29.2% of patients, acute grade 3 toxicity in 11.1%, and late grade 3 toxicity in 1.4% with no treatment-related deaths. CONCLUSIONS: SBRT appears to be a safe and effective means of salvaging recurrent, residual, or new primary NSCLC in or adjacent to a previous high-dose radiation field.

Stereotactic Body Radiation Therapy for Pulmonary Oligometastases Arising from Non-lung Primaries in Patients Without Extrapulmonary Disease.

Purpose Stereotactic body radiation therapy (SBRT) is increasingly used in the management of patients with oligometastatic cancers and is under prospective evaluation by the Radiation Therapy Oncology Group (RTOG). Here we report outcomes from a high-volume institution of patients treated with SBRT for pulmonary oligometastases. Materials and methods We conducted a retrospective review of 105 patients who had one to five pulmonary oligometastases (185 lesions) without extrapulmonary disease treated with SBRT from 2002-2014. Target failure-free survival (TFFS), progression-free survival (PFS), and overall survival (OS) were calculated. Univariate and multivariate Cox regression analyses were performed on factors predictive of outcomes. Results The median age at first SBRT was 68 years and the median follow-up was 29.5 months. The median time from initial diagnosis of primary to SBRT was 42.7 months; 14.3% had synchronous oligometastases and 76.7% had one to two pulmonary lesions at first SBRT. The distribution of primaries was as follows: 36.2% colorectal, 16.2% head/neck, 9.5% genitourinary, 9.5% sarcoma, 7.6% gynecologic, 6.7% other, 5.7% breast, 5% melanoma, and 4% esophageal. The median lesion size was 1.6 cm and the most common regimen was 60 Gy in three fractions (range: 12-60 Gy in one to five fractions). TFFS was 94.4% and 90.8% at two and three years, respectively. Two and three year OS were 87.9% and 60.2%, respectively. Median PFS and OS were 16.2 and 45.3 months, respectively. In multivariate analysis, age at primary cancer diagnosis and biologically effective dose with an alpha-beta ratio of 10 (BED10) were identified as factors significantly affecting OS (p<0.05). Conclusions Comprehensive treatment of pulmonary oligometastases with SBRT in the absence of extrapulmonary disease results in excellent target control and modest survival outcomes.

Single-institutional outcomes of adjuvant brachytherapy for Stage I endometrial cancer-Are outcomes consistent with randomized studies?

PURPOSE: Vaginal brachytherapy (VBT) alone has been shown to be a viable adjuvant treatment strategy for most patients with Stage I endometrioid endometrial cancer. We sought to examine our institutional data following practice pattern changes resulting from the publications of GOG-99 and PORTEC-2. METHODS AND MATERIALS: We retrospectively analyzed women who underwent adjuvant VBT after surgical staging for Stage 1 endometrioid endometrial cancer at our institution from 2007 to 2014. RESULTS: We identified 297 women. Median time to last followup or death was 52.3 months (interquartile range: 32.3-72.3 months). By International Federation of Gynecology and Obstetrics 2009 staging, 162 patients (54.5%) had Stage IA and 128 (43.1%) had Stage IB disease. Ninety-nine (33.3%) patients had Grade 1, 153 (51.5%) had Grade 2, and 45 (15.2%) had Grade 3 disease. According to GOG-249 and PORTEC-2 criteria, 167 (56.2%) and 127 (42.7%) patients were with high-intermediate-risk disease. Two women had Stage IB Grade 3 disease. The most common high-dose-rate-VBT regimen was 2100 cGy/three fractions to a depth of 5 mm. Four (two acute and two late) (1.3%) Grade 3 genitourinary toxicities were reported: three episodes of vaginal dehiscence (after second course of VBT, 2 months after completion of VBT, and 1 year after completion of VBT) and one episode of radiation necrosis. Twenty-one (7%) women recurred: three recurred in the vagina, two recurred in the pelvic lymph nodes, and 16 recurred distantly. CONCLUSIONS: Outcomes appear consistent with published randomized data in women with high-intermediate-risk endometrial cancer who are treated with brachytherapy alone. Recurrence and complication rates were minimal.

Stereotactic body radiation therapy for isolated hilar and mediastinal non-small cell lung cancers.

OBJECTIVES: The seminal phase II trial for pulmonary stereotactic body radiation therapy (SBRT) suggested that SBRT to central lesions resulted in unacceptable toxicity. Alternative dose-fractionation schemes have been proposed which may improve safety without compromise of efficacy. We report our institutional outcomes of SBRT for hilar/mediastinal non-small cell lung cancer (NSCLC). MATERIALS AND METHODS: A retrospective review was conducted of patients with NSCLC in a hilar or mediastinal nodal station which was treated with SBRT. Patients presented with a lesion involving the hilum or mediastinum from primary or oligorecurrent NSCLC. Kaplan-Meier with log-rank testing and Cox analysis were utilized for outcomes analysis. RESULTS: From 2008-2015, 40 patients with median age of 70 were treated with SBRT for primary/oligorecurrent hilar/mediastinal NSCLC with median follow-up of 16.4 months. 85% presented with oligorecurrent disease at a median of 22.4 months following definitive therapy. The aortico-pulmonary window was the target in 40%, the hilum in 25%, lower paratracheal in 20%, subcarinal in 10%, and prevascular in 5%. The median dose was 48Gy in 4 fractions (range: 35-48Gy in 4-5 fractions). Median overall (OS) and progression-free (PFS) survivals were 22.7 and 13.1 months, respectively. Two-year local control was 87.7% and not significantly different between hilar and mediastinal targets. Median PFS was significantly improved in patients with hilar vs mediastinal nodal targets: 33.3 vs 8.4 months, respectively (p=0.031). OS was not statistically different between hilar and mediastinal targets (p=0.359). On multivariable analysis, hilar vs mediastinal target predicted for PFS (HR 3.045 95%CI [1.044-8.833], p=0.042), as did shorter time to presentation in patients with oligorecurrence (HR 0.983 [95%CI 0.967-1.000], p=0.049). Acute grade 3+ morbidity was seen in 3 patients (hemoptysis, pericardial/pleural effusion, heart failure) and late grade 3+ morbidity (hemoptysis) in 1 patient. CONCLUSION: Hilar/mediastinal SBRT appears to be a safe technique for the local control of isolated nodal disease with limited toxicity from the fractionation schemes utilized.

Stereotactic body radiotherapy for locally-advanced unresectable pancreatic cancer-patterns of care and overall survival.

BACKGROUND: Unresectable pancreatic cancer remains a challenging disease to treat. Stereotactic body radiotherapy (SBRT) allows for a higher biologically equivalent dose in an abbreviated course more convenient for patients and the integration of systemic therapy. We sought to investigate utilization trends and survival outcomes for patients treated with pancreatic SBRT versus conventionally fractionated radiotherapy (CFRT). METHODS: We engaged the National Cancer Database (NCDB) from 1998-2012 and identified locally-advanced unresectable patients with histologically confirmed, non-metastatic, pancreatic adenocarcinoma who received radiotherapy. Patients who received CFRT (1.5-4.0 Gy per fraction to a dose of ≥45 Gy, n=11,879) were compared to those who received SBRT (6-15 Gy per fraction to a dose of ≥20 Gy, n=474). RESULTS: Median follow-up was 11.0 months (18.4 months for survivors). SBRT utilization increased from 0.2% to 7.4% from 1998 to 2012 (P<0.05). On multivariable analysis, factors predictive for preferential utilization of SBRT over CFRT were later year of diagnosis, age ≥75 years, increased facility volume, and no chemotherapy in the initial treatment plan. Unadjusted median survival was 11.2 months for CFRT vs. 12.6 months for SBRT (P=0.002). Results were consistent in the propensity matched model. Variables predictive for improved survival on multivariable analysis were diagnosis after 2010, younger age, lower comorbidity score, tumor size <3 cm, nodal stage zero, and receipt of chemotherapy (P<0.05). CONCLUSIONS: SBRT utilization has increased significantly and is associated with a small absolute improvement in overall survival (OS) compared to CFRT. The decreased treatment time, without apparent compromise in survival, makes SBRT an attractive option for patients with unresectable pancreatic cancer warranting further research.

Glioblastoma multiforme (GBM) in the elderly: initial treatment strategy and overall survival.

The EORTC trial which solidified the role of external beam radiotherapy (EBRT) plus temozolomide (TMZ) in the management of GBM excluded patients over age 70. Randomized studies of elderly patients showed that hypofractionated EBRT (HFRT) alone or TMZ alone was at least equivalent to conventionally fractionated EBRT (CFRT) alone. We sought to investigate the practice patterns and survival in elderly patients with GBM. We identified patients age 65-90 in the National Cancer Data Base (NCDB) with histologically confirmed GBM from 1998 to 2012 and known chemotherapy and radiotherapy status. We analyzed factors predicting treatment with EBRT alone vs. EBRT plus concurrent single-agent chemotherapy (CRT) using multivariable logistic regression. Similarly, within the EBRT alone cohort we compared CFRT (54-65 Gy at 1.7-2.1 Gy/fraction) to HFRT (34-60 Gy at 2.5-5 Gy/fraction). Multivariable Cox proportional hazards model (MVA) with propensity score adjustment was used to compare survival. A total of 38,862 patients were included. Initial treatments for 1998 versus 2012 were: EBRT alone = 50 versus 10%; CRT = 6 versus 50%; chemo alone = 1.6% (70% single-agent) versus 3.2% (94% single-agent). Among EBRT alone patients, use of HFRT (compared to CFRT) increased from 13 to 41%. Numerous factors predictive for utilization of CRT over EBRT alone and for HFRT over CFRT were identified. Median survival and 1-year overall survival were higher in the CRT versus EBRT alone group at 8.6 months vs. 5.1 months and 36.0 versus 15.7% (p < 0.0005 by log-rank, multivariable HR 0.65 [95% CI = 0.61-0.68, p < 0.0005], multivariable HR with propensity adjustment 0.66 [95% CI = 0.63-0.70, p < 0.0005]). For elderly GBM patients in the United States, CRT is the most common initial treatment and appears to offer a survival advantage over EBRT alone. Adoption of hypofractionation has increased over time but continues to be low.