The Emergence of Artificial Intelligence within Radiation Oncology Treatment Planning.

BACKGROUND: The future of artificial intelligence (AI) heralds unprecedented change for the field of radiation oncology. Commercial vendors and academic institutions have created AI tools for radiation oncology, but such tools have not yet been widely adopted into clinical practice. In addition, numerous discussions have prompted careful thoughts about AI's impact upon the future landscape of radiation oncology: How can we preserve innovation, creativity, and patient safety? When will AI-based tools be widely adopted into the clinic? Will the need for clinical staff be reduced? How will these devices and tools be developed and regulated? SUMMARY: In this work, we examine how deep learning, a rapidly emerging subset of AI, fits into the broader historical context of advancements made in radiation oncology and medical physics. In addition, we examine a representative set of deep learning-based tools that are being made available for use in external beam radiotherapy treatment planning and how these deep learning-based tools and other AI-based tools will impact members of the radiation treatment planning team. Key Messages: Compared to past transformative innovations explored in this article, such as the Monte Carlo method or intensity-modulated radiotherapy, the development and adoption of deep learning-based tools is occurring at faster rates and promises to transform practices of the radiation treatment planning team. However, accessibility to these tools will be determined by each clinic's access to the internet, web-based solutions, or high-performance computing hardware. As seen by the trends exhibited by many technologies, high dependence on new technology can result in harm should the product fail in an unexpected manner, be misused by the operator, or if the mitigation to an expected failure is not adequate. Thus, the need for developers and researchers to rigorously validate deep learning-based tools, for users to understand how to operate tools appropriately, and for professional bodies to develop guidelines for their use and maintenance is essential. Given that members of the radiation treatment planning team perform many tasks that are automatable, the use of deep learning-based tools, in combination with other automated treatment planning tools, may refocus tasks performed by the treatment planning team and may potentially reduce resource-related burdens for clinics with limited resources.

A risk assessment of automated treatment planning and recommendations for clinical deployment.

PURPOSE: To assess the risk of failure of a recently developed automated treatment planning tool, the radiation planning assistant (RPA), and to determine the reduction in these risks with implementation of a quality assurance (QA) program specifically designed for the RPA. METHODS: We used failure mode and effects analysis (FMEA) to assess the risk of the RPA. The steps involved in the workflow of planning a four-field box treatment of cervical cancer with the RPA were identified. Then, the potential failure modes at each step and their causes were identified and scored according to their likelihood of occurrence, severity, and likelihood of going undetected. Additionally, the impact of the components of the QA program on the detectability of the failure modes was assessed. The QA program was designed to supplement a clinic's standard QA processes and consisted of three components: (a) automatic, independent verification of the results of automated planning; (b) automatic comparison of treatment parameters to expected values; and (c) guided manual checks of the treatment plan. A risk priority number (RPN) was calculated for each potential failure mode with and without use of the QA program. RESULTS: In the RPA automated treatment planning workflow, we identified 68 potential failure modes with 113 causes. The average RPN was 91 without the QA program and 68 with the QA program (maximum RPNs were 504 and 315, respectively). The reduction in RPN was due to an improvement in the likelihood of detecting failures, resulting in lower detectability scores. The top-ranked failure modes included incorrect identification of the marked isocenter, inappropriate beam aperture definition, incorrect entry of the prescription into the RPA plan directive, and lack of a comprehensive plan review by the physician. CONCLUSIONS: Using FMEA, we assessed the risks in the clinical deployment of an automated treatment planning workflow and showed that a specialized QA program for the RPA, which included automatic QA techniques, improved the detectability of failures, reducing this risk. However, some residual risks persisted, which were similar to those found in manual treatment planning, and human error remained a major cause of potential failures. Through the risk analysis process, we identified three key aspects of safe deployment of automated planning: (a) user training on potential failure modes; (b) comprehensive manual plan review by physicians and physicists; and (c) automated QA of the treatment plan.

Cost comparison of treatment for oropharyngeal carcinoma.

OBJECTIVES/HYPOTHESIS: Based on current guidelines, surgical and nonsurgical therapies are viable frontline treatment for patients with locoregional oropharyngeal carcinoma (OPC). We sought to compare financial parameters between chemoradiation and transoral robotic surgery (TORS) in this patient population. STUDY DESIGN: Case-control study. METHODS: In this study we identified patients with selected American Joint Committee on Cancer 7th Edition stage II to IVa OPC treated with TORS between January 2013 and December 2014. Fifteen patients who underwent TORS were stage matched with 15 patients treated with chemoradiation. Total charges and cost data for each patient were analyzed at 4-month and 1-year time points; functional and oncologic outcomes were assessed. RESULTS: There were no significant differences in functional and oncologic outcomes. Patients undergoing TORS had a longer inpatient hospital stay, and most required a nasogastric tube for an average of 3.5 days. There were no local or regional recurrences. Across all time points, the TORS group had lower charges and costs compared to the chemoradiation group, with 14% lower costs at 1 year. In the chemoradiation group, nearly two-thirds of costs came from radiation therapy and pharmacy expenses. Chemotherapy accounted for most pharmacy costs. The costs of operating the surgical robot accounted for a about half of surgical costs. CONCLUSIONS: Selected patients with stage II to IVa oropharyngeal carcinoma treated with TORS may incur lower costs than those treated nonsurgically. With rising healthcare spending, the financial impact of treatment might be considered for those patients eligible for treatment regimens with comparable functional and oncologic outcomes. LEVEL OF EVIDENCE: 3b Laryngoscope, 129:1604-1609, 2019.

Survival of patients with head and neck cancer treated with definitive radiotherapy and concurrent cisplatin or concurrent cetuximab: A Surveillance, Epidemiology, and End Results-Medicare analysis.

BACKGROUND: Cisplatin and cetuximab are both systemic therapies commonly used in combination with radiation (RT) for the definitive treatment of head and neck cancers, but their comparative efficacy is unclear. METHODS: Patients with locoregionally advanced (American Joint Committee on Cancer stage III-IVB) squamous cell carcinomas of the oropharynx, larynx, or hypopharynx were identified in the Surveillance, Epidemiology, and End Results-Medicare database. Patients received either cisplatin or cetuximab concurrent with RT, as determined by Medicare claims. The primary study outcome was head and neck cancer-specific mortality (CSM) analyzed with competing risks. Filtering, propensity score matching, and multivariable Fine-Gray regression were used to adjust for differences between the cisplatin and cetuximab cohorts, including age, comorbidity, and cycles of systemic therapy received. RESULTS: The total cohort consisted of 1395 patients, of whom 786 (56%) received cisplatin and 609 (44%) received cetuximab; the median follow-up was 3.5 years in the patients who remained alive. In the cetuximab cohort, CSM was significantly higher than in the cisplatin cohort (39% vs 25% at 3 years; P < .0001). In the matched cohorts (n = 414), the adjusted hazard ratio of CSM for cetuximab was 1.65 (95% confidence interval, 1.30-2.09; P < .0001) relative to cisplatin, corresponding to an absolute difference of approximately 10% in both CSM and overall survival at 3 years. Cetuximab was associated with less dysphagia, more dermatitis, and a similar incidence of mucositis. CONCLUSIONS: In this sizeable, national patient population, treatment with cetuximab was associated with significantly higher CSM than cisplatin. These results suggest that cisplatin may be the preferred chemotherapeutic agent in this setting.

Development and Feasibility of Bundled Payments for the Multidisciplinary Treatment of Head and Neck Cancer: A Pilot Program.

PURPOSE: Despite growing interest in bundled payments to reduce the costs of care, this payment method remains largely untested in cancer. This 3-year pilot tested the feasibility of a 1-year bundled payment for the multidisciplinary treatment of head and neck cancers. METHODS: Four prospective treatment-based bundles were developed for patients with selected head and neck cancers. These risk-adjusted bundles covered 1 year of care that began with primary cancer treatment. Manual processes were developed for patient identification, enrollment, billing, and payment. Patients were prospectively identified and enrolled, and bundled payments were made at treatment start. Operational metrics tracked incremental effort for pilot processes and average payment cycle time compared with fee-for-service (FFS) payments. RESULTS: This pilot confirmed the feasibility of a 1-year prospective bundled payment for head and neck cancers. Between November 2014 and October 2016, 88 patients were enrolled successfully with prospective bundled payments. Through September 2017, 94% of patients completed the pilot with 6% still enrolled. Manual pilot processes required more effort than anticipated; claims processing was the most time-consuming activity. The production of a bundle bill took an additional 15 minutes versus FFS billing. The average payment cycle time was 37 days (range, 15 to 141 days) compared with a 15-day average under FFS. CONCLUSION: Prospective bundled payments were successfully implemented in this pilot. Additional pilots should study this payment method in higher-volume cancers. Robust systems are needed to automate patient identification, enrollment, billing, and payment along with policies that reduce administrative burden and allow for the introduction of novel cancer therapies.

Reduced feeding tube duration with intensity-modulated radiation therapy for head and neck cancer: A Surveillance, Epidemiology, and End Results-Medicare Analysis.

BACKGROUND: Intensity-modulated radiation therapy (IMRT) is a technologically advanced and resource-intensive method of delivering radiation therapy (RT) and is used to minimize toxicity for patients with head and neck cancer (HNC). Dependence on feeding tubes is a significant marker of toxicity of RT. The objective of this analysis was to compare the placement and duration of feeding tube use among patients with HNC from 1999 through 2011. METHODS: The cohort, demographics, and cancer-related variables were determined using the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database, and claims data were used to analyze treatment details. RESULTS: In total, 2993 patients were identified. At a median follow-up of 47 months, 54.4% of patients had ever had a feeding tube placed. The median duration from feeding tube placement to removal was 277 days. On zero-inflated negative binomial regression, patients who received IMRT and 3-dimensional RT (3DRT) (non-IMRT) had similar rates of feeding tube placement (odds ratio, 1.10; P = .35); however, patients who received 3DRT had a feeding tube in place 1.18 times longer than those who received IMRT (P = .03). The difference was only observed among patients who received definitive RT; patients who underwent surgery and also received adjuvant RT had no statistically significant difference in feeding tube placement or duration. CONCLUSIONS: Patients with HNC who received definitive IMRT had a significantly shorter duration of feeding tube placement than those who received 3DRT. These data suggest that there may be significant quality-of-life benefits to IMRT with respect to long-term swallowing function in patients with HNC. Cancer 2017;123:283-293. © 2016 American Cancer Society.

Favorable patient reported outcomes following IMRT for early carcinomas of the tonsillar fossa: Results from a symptom assessment study.

BACKGROUND: A questionnaire-based study was conducted to assess long-term patient reported outcomes (PROs) following definitive IMRT-based treatment for early stage carcinomas of the tonsillar fossa. METHODS: Participants had received IMRT with or without systemic therapy for squamous carcinoma of the tonsillar fossa (T1-2 and N0-2b) with a minimum follow-up of 2years. Patients completed a validated head and neck cancer-specific PRO instrument, the MD Anderson Symptom Inventory-Head and Neck module (MDASI-HN). Symptoms were compared between treatment groups of interest and overall symptom burden was evaluated. RESULTS: Of 139 participants analyzed, 51% had received ipsilateral neck IMRT, and 62% single modality IMRT alone (no systemic therapy). There were no differences in mean severity ratings for the top-ranked individual symptoms or symptom interference for those treated with bilateral versus ipsilateral neck IMRT alone. However, 40% of those treated with bilateral versus 25% of those treated with ipsilateral neck RT alone reported moderate-to-severe levels of dry mouth (p=0.03). Fatigue, numbness/tingling, and constipation were rated more severe for those who had received systemic therapy (p<0.05 for each), but absolute differences were small. Overall, 51% had no more than mild symptom ratings across all 22 symptoms assessed. CONCLUSIONS: The long-term patient reported symptom profile in this cohort of tonsil cancer survivors treated with definitive IMRT-based treatment showed a majority of patients with no more than mild symptoms, low symptom interference, and provides an opportunity for future comparison studies with other treatment approaches.

Beam path toxicity in candidate organs-at-risk: assessment of radiation emetogenesis for patients receiving head and neck intensity modulated radiotherapy.

BACKGROUND: To investigate potential dose-response relationship between radiation-associated nausea and vomiting (RANV) reported during radiotherapy and candidate nausea/vomiting-associated regions of interest (CNV-ROIs) in head and neck (HNC) squamous cell carcinomas. METHODS AND MATERIAL: A total of 130 patients treated with IMRT with squamous cell carcinomas of head and neck were evaluated. For each patient, CNV-ROIs were segmented manually on planning CT images. Clinical on-treatment RANV data were reconstructed by a review of the records for all patients. Dosimetric data parameters were recorded from dose-volume histograms. Nausea and vomiting reports were concatenated as a single binary "Any N/V" variable, and as a "CTC-V2+" variable. RESULTS: The mean dose to CNV-ROIs was higher for patients experiencing RANV events. For patients receiving IMRT alone, a dose-response effect was observed with varying degrees of magnitude, at a statistically significant level for the area postrema, brainstem, dorsal vagal complex, medulla oblongata, solitary nucleus, oropharyngeal mucosa and whole brain CNV-ROIs. CONCLUSION: RANV is a common therapy-related morbidity facing patients receiving HNC radiotherapy, and, for those receiving radiotherapy-alone, is associated with modifiable dose to specific CNS structures.

Improved survival using intensity-modulated radiation therapy in head and neck cancers: a SEER-Medicare analysis.

BACKGROUND: Intensity-modulated radiation therapy (IMRT) is a technologically advanced, and more expensive, method of delivering radiation therapy with a goal of minimizing toxicity. It has been widely adopted for head and neck cancers; however, its comparative impact on cancer control and survival remains unknown. The goal of this analysis was to compare the cause-specific survival (CSS) for patients with head and neck cancers treated with IMRT versus non-IMRT from 1999 to 2007. METHODS: CSS was determined using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database and analyzed regarding treatment details, including the use of IMRT versus non-IMRT, using claims data. Hazard ratios (HRs) were estimated by the frailty model with a propensity score matching cohort and instrumental variable analysis. RESULTS: A total of 3172 patients were identified. With a median follow-up of 40 months, patients treated with IMRT had a statistically significant improvement in CSS compared with those treated with non-IMRT (84.1% versus 66.0%; P < .001). When each anatomic subsite was analyzed separately, all respective subgroups of patients treated with IMRT had better CSS than those treated with non-IMRT. In multivariable survival analyses, patients treated with IMRT were associated with better CSS (HR = 0.72, 95% confidence interval = 0.59 to 0.90 for propensity score matching; HR = 0.60, 95% confidence interval = 0.41 to 0.88 for instrumental variable analysis). CONCLUSIONS: Patients with head and neck cancers who were treated with IMRT experienced significant improvements in CSS compared with patients treated with non-IMRT techniques. This suggests there may be benefits to IMRT in cancer outcomes, in addition to toxicity reduction, for this patient population.

Evaluating the impact of patient, tumor, and treatment characteristics on the development of jaw complications in patients treated for oral cancers: a SEER-Medicare analysis.

BACKGROUND: Jaw complications, including osteoradionecrosis, are significant sequelae of radiation therapy (RT) for oral cancers. This study identifies the impact of patient, tumor, and treatment characteristics on the development of jaw complications in patients treated with RT. METHODS: The Surveillance, Epidemiology, and End Results (SEER)-Medicare database was used to identify patients treated with RT for oral cancers from 1999 to 2007. Jaw complications were identified by International Classification of Diseases 9th revision (ICD-9) diagnosis codes and/or related procedures using Current Procedural Terminology (CPT) and ICD-9 codes. RESULTS: A total of 1848 patients were identified. With a median follow-up of 2.5 years, 297 patients (16.1%) developed jaw complications: 226 patients had a diagnosis, 41 patients had a procedure, and 30 patients had both. On multivariate analysis, female sex, lack of chemotherapy use, and fewer comorbidities were associated with a statistically significant increase in jaw complications. CONCLUSIONS: Even with modern techniques, jaw complications are a notable and potentially devastating side effect of RT for oral cancers.

Prospective assessment of an atlas-based intervention combined with real-time software feedback in contouring lymph node levels and organs-at-risk in the head and neck: Quantitative assessment of conformance to expert delineation.

PURPOSE: A number of studies have previously assessed the role of teaching interventions to improve organ-at-risk (OAR) delineation. We present a preliminary study demonstrating the benefit of a combined atlas and real time software-based feedback intervention to aid in contouring of OARs in the head and neck. METHODS AND MATERIALS: The study consisted of a baseline evaluation, a real-time feedback intervention, atlas presentation, and a follow-up evaluation. At baseline evaluation, 8 resident observers contoured 26 OARs on a computed tomography scan without intervention or aid. They then received feedback comparing their contours both statistically and graphically to a set of atlas-based expert contours. Additionally, they received access to an atlas to contour these structures. The resident observers were then asked to contour the same 26 OARs on a separate computed tomography scan with atlas access. In addition, 6 experts (5 radiation oncologists specializing in the head and neck, and 1 neuroradiologist) contoured the 26 OARs on both scans. A simultaneous truth and performance level estimation (STAPLE) composite of the expert contours was used as a gold-standard set for analysis of OAR contouring. RESULTS: Of the 8 resident observers who initially participated in the study, 7 completed both phases of the study. Dice similarity coefficients were calculated for each user-drawn structure relative to the expert STAPLE composite for each structure. Mean dice similarity coefficients across all structures increased between phase 1 and phase 2 for each resident observer, demonstrating a statistically significant improvement in overall OAR-contouring ability (P < .01). Additionally, intervention improved contouring in 16/26 delineated organs-at-risk across resident observers at a statistically significant level (P ≤ .05) including all otic structures and suprahyoid lymph node levels of the head and neck. CONCLUSIONS: Our data suggest that a combined atlas and real-time feedback-based educational intervention detectably improves contouring of OARs in the head and neck.