Experienced Radiation Oncologist Research Articles

Patterns of treatment failure in patients with sinonasal squamous cell carcinoma after chemoradiotherapy.

To investigate the failure patterns based on precision radiation treatment and to determine the predictive factors of treatment failure for sinonasal squamous cell carcinoma (SNSCC) patients. This was a retrospective study that included 214 cases of treatment failure from 441 consecutive patients. Two experienced radiation oncologists evaluated the tumor volume of cases with local recurrence. The 5-year overall survival (OS), progression-free survival rates (PFS), and distant-metastasis-free survival (DMFS) were estimated. Investigations were performed on the factors that predicted local failure or distant metastasis. 140 (31.7%) patients developed local recurrence, 24 (5.4%) experienced regional failure, and 65 (14.7%) underwent distant metastasis. In-field, marginal, and out-of-field failures occurred in 55.7% (78/140), 33.6% (47/140), and 10.7% (15/140) of patients with local recurrence, respectively. In logistic regression analysis, factors statistically significant for total local failure included treatment mode (p < 0.01), chemotherapy (p < 0.01), and surgical margins (p < 0.01). Primary tumors with poor differentiation (p = 0.018) and R2 resection margin (p = 0.009) were more prone to develop distant failure. The 5-year OS, PFS, and DMFS rates were 57.8%, 52.0%, and 56.7% for the whole cohort. In univariate and multivariate analysis, the skull base involvement was an independent predictor for poorer OS and PFS; orbital invasion was an independent predictor for poorer OS. Local recurrence and distant metastasis were the most common failure modes. Treatment mode, chemotherapy, and surgical margins were related to local recurrence. Poor differentiation and R2 resection margin were predictors for distant failure. Local recurrence is the most common failure pattern in patients with sinonasal squamous cell carcinoma who accepted chemoradiotherapy, and marginal and out-of-field failures occurred in 44.3% of patients with local recurrence.

Deep learning-based segmentation for high-dose-rate brachytherapy in cervical cancer using 3D Prompt-ResUNet.

To develop and evaluate a 3D Prompt-ResUNet module that utilized the prompt-based model combined with 3D nnUNet for rapid and consistent autosegmentation of high-risk clinical target volume and organ at risk in high-dose-rate brachytherapy for cervical cancer patients. &#xD;Approach. We used 73 computed tomography (CT) and 62 magnetic resonance imaging (MRI) scans from 135 (103 for training, 16 for validation, and 16 for testing) cervical cancer patients across two hospitals for HRCTV and OAR segmentation. A novel comparison of the deep learning neural networks 3D Prompt-ResUNet, nnUNet, and SAM-Med3D was applied for the segmentation. Evaluation was conducted in two parts: geometric and clinical assessments. Quantitative metrics included the Dice similarity coefficient (DSC), 95th percentile Hausdorff distance (HD95%), Jaccard index (JI), and Matthews correlation coefficient (MCC). Clinical evaluation involved interobserver comparison, 4-grade expert scoring, and a double-blinded Turing test.&#xD;Main results. The Prompt-ResUNet model performed most similarly to experienced radiation oncologists, outperforming less experienced ones. During testing, the DSC, HD95% (mm), JI, and MCC value (mean ± SD) for HRCTV were 0.92±0.03, 2.91 ± 0.69, 0.85± 0.04, and 0.92 ± 0.02, respectively. For the bladder, these values were 0.93 ± 0.05, 3.07 ± 1.05, 0.87 ± 0.08, and 0.93 ± 0.05, respectively. For the rectum, they were 0.87 ± 0.03, 3.54 ± 1.46, 0.78 ± 0.05, and 0.87 ± 0.03, respectively. For the sigmoid, they were 0.76 ± 0.11, 7.54 ± 5.54, 0.63 ± 0.14, and 0.78 ± 0.09, respectively. The Prompt-ResUNet achieved a clinical viability score of at least 2 in all evaluation cases (100%) for both HRCTV and bladder and exceeded the 30% positive rate benchmark for all evaluated structures in the Turing test.&#xD;Significance. The Prompt-ResUNet architecture demonstrated high consistency with ground truth (GT) in autosegmentation of HRCTV and OARs, reducing interobserver variability and shortening treatment times.&#xD.

Cardiac substructure radiation dose study of conventional radiotherapy for stage N2-3 non-small cell lung cancer.

e20071 Background: There has been widespread reporting on predictive dose parameters for cardiac substructures in radiotherapy for non-small cell lung cancer (NSCLC). However, most advanced NSCLC patients are staged as N2-3, and there is a lack of research on the dose distribution, dose contribution and prognosis of the heart in conventional radiotherapy for N2-3 stage NSCLC. Methods: Retrospective analysis of 274 patients of stage N2-3 NSCLC treated with radiotherapy (RT) of prescription dose from 60Gy to 70Gy. Experienced radiation oncologists delineated the cardiac substructures. All patients were categorized according to different lymph node irradiation and tumor spatial locations. Spearman and linear regression were used for variation of dose-volume parameters, as well as the impact of different primary tumor locations and irradiated lymph nodes on the parameters of the entire heart and substructures. Receiver operating characteristic (ROC) curves, Kaplan-Meier, and Cox regression were performed on the influencing factors of adverse cardiac events after radiotherapy and prognosis analysis. Results: There are significant differences in the dose-volume parameters of the whole heart and its substructures. Mean heart doses (MHD),and the mean doses (Dmean) for the left atrium, right atrium, right ventricle, left ventricle and the left anterior descending coronary artery (LAD) were 15.7Gy (range 0.2-30.0Gy), 19.7Gy (range 0.3-55.0Gy), 9.7Gy (range 0.3-59.8Gy), 7.9Gy (range 0.1-48.2Gy), 6.4Gy (range 0.1-52.6Gy) and 5.2Gy (range 0.1-45.0Gy), respectively. The left atrium Dmean has the most significant impact on MHD compared to other substructures (B = 0.418, P &lt; 0.05). Dose-volume parameters of the whole heart and the left atrium were positively correlated with the irradiation of the 7th lymph node region (P &lt; 0.05). MHD and left atrium Dmean for tumors located in the lower region were significantly higher than those in the upper region (P &lt; 0.001). Patients with LAD V30 ≥ 26.9% had a significantly higher risk of mortality compared to those with LAD V30 &lt; 26.9% (P = 0.003, HR = 2.883, 95% CI = 1.436-5.788). Conclusions: With the dose distribution differences among the whole heart and its substructures. We observed the base of the heart received the highest dose, the dose to the left atrium is particularly required attention. The irradiation of different lymph nodes and different primary lesions has an impact on the dose variations of different cardiac structures. LAD V30 is an independent influencing factor of overall survival (OS) for N2-3 stage NSCLC. This study provides considerable information for the dose optimization of cardiac substructure in radiotherapy planning for N2-3 stage NSCLC.

MRI-Linac radiation therapy outcomes in patients with prostate, breast and other advanced malignancies.

e13811 Background: MR-guided radiation therapy offers an innovative approach, providing superior soft tissue contrast that enables clinicians to deliver higher doses to the target site with fewer fractions. The objective of the study is to assess objective response rates and adverse effects related to a community-based single-institution experience with the ViewRay MRIdian. Methods: An IRB-approved registrational trial was conducted at the Providence Cancer Institute starting May 2020, with data cutoff of September 2023. All patients treated underwent MR screening, and the target site(s) were treated using hypofractionated radiation therapy in ≤5 fractions. The primary endpoint was to assess adverse effects attributed to radiation therapy. Secondary measures included local recurrence-free survival (LRFS), distant metastasis-free survival (DMFS), overall survival (OS), and change in prostate-specific antigen (PSA) for patients treated definitively for prostate cancer. The assessments were performed under the guidance of experienced radiation oncologists. Results: From May 2020 to September 2023, n=257 patients received treatment on the MRIdian with a median follow-up of 430 days [range 4-1221 days]. One and 2-year LRFS was 92% and 87%, and excluding prostate and breast cancer, 83% and 70%. Definitive prostatestereotactic body radiation therapy(SBRT) was delivered to 117 patients and significantly reduced PSA (mean difference -6.27, p&lt;0.0001), with no local failures to date. One and 2-year DMFS was 75% and 67%, though if excluding definitive prostate SBRT and partial breast irradiation, these were 48% and 33%. The OS rate at 1 and 2 years was 86% and 76%. Excluding prostate and breast cancer, the 1 and 2-year OS was 74% and 55%. Grade 3+ adverse events possibly or definitely attributable to radiation therapy were noted in 5% (n=13/257) of patients, including 2 possible treatment-associated deaths (bowel perforation following liver SBRT, and respiratory failure following lung SBRT and trastuzumab-deruxtecan systemic therapy). Conclusions: Our institutional experience with the MRIdian demonstrated high levels of local control, while overall survival was driven by out-of-field failures. Serious adverse events were infrequent, but more common in abdominal targets highlighting the need for careful planning and tight organs at risk (OAR) constraints, particularly if systemic therapy is delivered soon after radiation.

Stereotactic Body Radiation Therapy for Clinically Localized Prostate Cancer in Men With Hip Prostheses: A Cautionary Note.

Stereotactic body radiation therapy (SBRT) has been established as a safe and effective treatment for prostate cancer. SBRT requires high accuracy to reduce treatment margins. Metal hip prostheses create artifacts that distort pelvic imaging and potentially decrease the accuracy of target/organ at risk (OAR) identification and radiation dose calculations. Data on the safety and efficacy of SBRT after hip replacement is limited. This single-institution study sought to evaluate the safety and local control following SBRT for prostate cancer in men with hip replacements. 23 patients treated with localized prostate cancer and a history of pre-treatment hip replacement, treated with SBRT from 2007 to 2017 at MedStar Georgetown University Hospital were included in this retrospective analysis. Treatment was administered with the CyberKnife® (Accuray Incorporated, Sunnyvale, CA) at doses of 35 Gy or 36.25 Gy in 5 fractions. The targets and OARs were identified and contoured by a single experienced Radiation Oncologist (SPC).The adequacy of the CT and T2W MRI images for treatment planning was assessed with a three-point scale (good, adequate, or suboptimal).During treatment planning, care was taken to avoid treatment beams that directly traversed the hip prosthesis. Toxicities were recorded and scored using the Common Terminology Criteria for Adverse Events version 4.0 (CTCAE v.4.0). Local recurrence was confirmed by magnetic resonance imaging and/or prostate biopsy. The median follow-up was seven years. The patients were elderly (median age = 71 years) with a high rate of comorbidities (Charlson Comorbidity Index > 2 in 25%). Four patients had bilateral hip replacements. The majority of patients were low to intermediate risk per the D'Amico classification. Around 13% received upfront ADT. In total, 13 patients were treated with 35 Gy, and 10 were treated with 36.25 Gy.The rates of late > Grade 3 GU toxicity and> Grade 2 GI toxicity were 8.6% and 4.3%, respectively. There were no Grade 4 or 5 toxicities.Six patients (26%) developed a local recurrence at a median time of 7.5 years. Of these six patients, four had unilateral hip replacements and two had bilateral.Three underwent salvage cryotherapy and three received salvage ADT. In the general population, high-grade toxicities and local recurrences are uncommon following prostate SBRT.However, in this cohort of patients with prior hip replacements, prostate SBRT had higher than expected rates of late toxicity and local recurrence. In the opinion of the authors, such patients should be counseled regarding an elevated risk of late toxicity and local recurrence with prostate SBRT. With its ultrasound guidance, brachytherapy would have the advantage of circumventing the need for MRI/CT-based imaging and thus may represent a preferable radiation alternative in this patient population. If these patients are treated with SBRT, they should be monitored closely for local recurrence so early salvage can be performed. We hopethat recent advances in metal artifact reduction techniques and dose-calculation algorithms will improve future outcomes.

A deep learning-based 3D Prompt-nnUnet model for automatic segmentation in brachytherapy of postoperative endometrial carcinoma.

To create and evaluate a three-dimensional (3D) Prompt-nnUnet module that utilizes the prompts-based model combined with 3D nnUnet for producing the rapid and consistent autosegmentation of high-risk clinical target volume (HR CTV) and organ at risk (OAR) in high-dose-rate brachytherapy (HDR BT) for patients with postoperative endometrial carcinoma (EC). On two experimental batches, a total of 321 computed tomography (CT) scans were obtained for HR CTV segmentation from 321 patients with EC, and 125 CT scans for OARs segmentation from 125 patients. The numbers of training/validation/test were 257/32/32 and 87/13/25 for HR CTV and OARs respectively. A novel comparison of the deep learning neural network 3D Prompt-nnUnet and 3D nnUnet was applied for HR CTV and OARs segmentation. Three-fold cross validation and several quantitative metrics were employed, including Dice similarity coefficient (DSC), Hausdorff distance (HD), 95th percentile of Hausdorff distance (HD95%), and intersection over union (IoU). The Prompt-nnUnet included two forms of parameters Predict-Prompt (PP) and Label-Prompt (LP), with the LP performing most similarly to the experienced radiation oncologist and outperforming the less experienced ones. During the testing phase, the mean DSC values for the LP were 0.96±0.02, 0.91±0.02, and 0.83±0.07 for HR CTV, rectum and urethra, respectively. The mean HD values (mm) were 2.73±0.95, 8.18±4.84, and 2.11±0.50, respectively. The mean HD95% values (mm) were 1.66±1.11, 3.07±0.94, and 1.35±0.55, respectively. The mean IoUs were 0.92±0.04, 0.84±0.03, and 0.71±0.09, respectively. A delineation time<2.35s per structure in the new model was observed, which was available to save clinician time. The Prompt-nnUnet architecture, particularly the LP, was highly consistent with ground truth (GT) in HR CTV or OAR autosegmentation, reducing interobserver variability and shortening treatment time.

Evaluation of a deep image-to-image network (DI2IN) auto-segmentation algorithm across a network of cancer centers.

Due to manual OAR contouring challenges, various automatic contouring solutions have been introduced. Historically, common clinical auto-segmentation algorithms used were atlas-based, which required maintaining a library of self-made contours. Searching the collection was computationally intensive and could take several minutes to complete. Deep learning approaches have shown significant benefits compared to atlas-based methods in improving segmentation accuracy and efficiency in auto-segmentation algorithms. This work represents the first multi-institutional study to describe and evaluate an AI algorithm for the auto-segmentation of organs at risk (OARs) based on a deep image-to-image network (DI2IN). The AI-Rad Companion Organs RT (AIRC) algorithm (Siemens Healthineers, Erlangen, Germany) uses a two-step approach for segmentation. In the first step, the target organ region in the optimal input image is extracted using a trained deep reinforcement learning network (DRL), which is then used as input to create the contours in the second step based on DI2IN. The study was initially designed as a prospective single-center evaluation. The automated contours generated by AIRC were evaluated by three experienced board-certified radiation oncologists using a four-point scale where 4 is clinically usable and 1 requires re-contouring. After seeing favorable results in a single-center pilot study, we decided to expand the study to six additional institutions, encompassing eight additional evaluators for a total of 11 physician evaluators across seven institutions. One hundred and fifty-six patients and 1366 contours were prospectively evaluated. The five most commonly contoured organs were the lung (136 contours, average rating = 4.0), spinal cord (106 contours, average rating = 3.1), eye globe (80 contours, average rating = 3.9), lens (77 contours, average rating = 3.9), and optic nerve (75 contours, average rating = 4.0). The average rating per evaluator per contour was 3.6. On average, 124 contours were evaluated by each evaluator. 65% of the contours were rated as 4, and 31% were rated as 3. Only 4% of contours were rated as 1 or 2. Thirty-three organs were evaluated in the study, with 19 structures having a 3.5 or above average rating (ribs, abdominopelvic cavity, skeleton, larynx, lung, aorta, brachial plexus, lens, eye globe, glottis, heart, parotid glands, bladder, kidneys, supraglottic larynx, submandibular glands, esophagus, optic nerve, oral cavity) and the remaining organs having a rating of 3.0 or greater (female breast, proximal femur, seminal vesicles, rectum, sternum, brainstem, prostate, brain, lips, mandible, liver, optic chiasm, spinal cord, spleen). No organ had an average rating below 3. AIRC performed well with greater than 95% of contours accepted by treating physicians with no or minor edits. It supported a fully automated workflow with the potential for time savings and increased standardization with the use of AI-powered algorithms for high-quality OAR contouring.

Comparison between pelvic MRI, CT, and PET/CT in baseline staging and radiation planning of anal squamous cell carcinoma.

To investigate the differences in baseline staging of anal squamous cell carcinoma based on CT, MRI, and PET/CT, and the resultant impact on the radiation plan. This retrospective study included consecutive patients with anal squamous cell carcinoma who underwent baseline pelvic MRI, CT, and PET/CT (all examinations within 3weeks of each other) from January 2010 to April 2020. CTs, MRIs, and PET/CTs were re-interpreted by three separate radiologists. Several imaging features were assessed; tumor stage was determined based on the eight edition of the American Joint Committee on Cancer (AJCC) staging manual; and T (tumor), N (node), and M (metastasis) categories were determined based on National Comprehensive Cancer Network (NCCN) guidelines. Radiologist assessments were then randomly presented to a radiation oncologist who formulated the radiation plan in a blinded fashion. Across 28 patients (median age, 62years [range, 31-78], T-category classification was significantly different on PET/CT compared to MRI and CT (p = 0.037 and 0.031, respectively). PET/CT staged a higher proportion of patients with T1/T2 disease (16/28, 57%) compared to MRI (11/28, 39%) and CT (10/28, 36%). MRI staged a higher proportion of patients with T3/T4 disease (14/28, 50%) compared to CT (12/28, 43%) and PET/CT (11/28, 39%). However, there was no significant difference between the three imaging modalities in terms of either N-category, AJCC staging, or NCCN TNM group classification, or in treatment planning. Our exploratory study showed that MRI demonstrated a higher proportion of T3/T4 tumors, while PET/CT demonstrated more T1/T2 tumors; however, MRI, CT, and PET/CT did not show any significant differences in AJCC and TNM group categories, nor was there any significant difference in treatment doses between them when assessed independently by an experienced radiation oncologist.

Comparing breath hold versus free breathing irradiation for left-sided breast radiotherapy by PlanIQ™

BackgroundBreast cancer is the most widespread cancer in women and young women worldwide. Moving towards customised radiotherapy, balancing the use of the available technology with the best treatment modality may not be an easy task in the daily routine. This study aims to evaluate the effectiveness of introducing IQ-feasibility into clinical practice to support the decision of free-breathing (FB) versus breath-hold (BH) left-sided breast irradiations, in order to optimise the technology available and the effectiveness of the treatment.MethodsThirty-five patients who received 3D radiotherapy treatment of the left breast in deep-inspiration BH were included in this retrospective study. Computed tomography scans in FB and BH were acquired for each patient; targets contoured in both imaging datasets by an experienced radiation oncologist, and organs at risk delineated using automatic segmentation software were exported to PlanIQ™ (Sun Nuclear Corp.) to generate feasibility dose volume histogram (FDVHs). The dosimetric parameter of BH versus FB FDVH, and BH clinical dataset versus BH FDVH were compared.ResultsA total of 30 patients out of 35 patients analysed, presented for the BH treatments a significant reduction (p < 0.05) in the heart mean dose (Dm\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${{\ ext{D}}}_{{\ ext{m}}}$$\\end{document}), volume receiving 5 Gy (V5Gy\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${{\ ext{V}}}_{5{\ ext{Gy}}}$$\\end{document}) and 20 Gy (V20Gy\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${{\ ext{V}}}_{20{\ ext{Gy}}}$$\\end{document}), of 35.7%, 54.5%, and 2.1%, respectively; for the left lung, a lower reduction was registered and significant only for V5Gy\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${{\ ext{V}}}_{5{\ ext{Gy}}}$$\\end{document} (21.4%, p = 0.046). For the remaining five patients, the FDVH cut-off points of heart and lung were superimposable with differences of less than 1%. Heart and left lung dosimetric parameters of the BH clinical plans are located in the difficult zone of the FDVH and differ significantly (p < 0.05) from the corresponding parameters of the FDVH curves delimiting this buffer area between the impossible and feasible zones, respectively.ConclusionThe use of PlanIQTM as a decision-support tool for the FB versus BH treatment delivery modality allows customisation of the treatment technique using the most appropriate technology for each patient enabling accurate management of available technologies.

Artificial Intelligence–Based Autosegmentation: Advantages in Delineation, Absorbed Dose-Distribution, and Logistics

PurposeTo compare the performance of artificial intelligence (AI) in auto-contouring compared to a human practitioner in terms of precision, differences in dose distribution and time consumption. Methods and MaterialsDatasets of previously irradiated patients in three different segments (head and neck- (HNC), breast- (BC) and prostate cancer (PC)) were retrospectively collected. An experienced radiation oncologist (MD) performed organs-at-risk (OARs) and standard clinical target volume (CTV) delineations as baseline structures for comparison. AI-based auto-contours were generated in two additional CT copies; therefore, three groups were assessed: MD alone, AI alone, AI plus MD corrections (AI+C). Differences in Dice similarity coefficient (DSC) and person-hour burden were assessed. Furthermore, changes in clinically relevant dose-volume parameters were evaluated and compared. ResultsSeventy-five previously treated cases were collected (25 per segment) for the analysis. When compared to MD contours, the mean DSC scores were higher than 0.7 for 74% and 80% of AI and AI+C, respectively. After corrections, 17.1% structures presented DSC score deviations higher than 0.1 and 10.4% dose-volume parameters significantly changed in AI-contoured structures. The time consumption assessment yielded mean person-hour reductions of 68%, 51% and 71% for BC, PC and HNC, respectively. ConclusionIn great extent, AI yielded clinically acceptable OARs and certain CTVs in the explored anatomical segments. Sparse correction and assessment requirements place AI+C as a standard workflow. Minimal clinically relevant differences in OAR-exposure were identified. A substantial amount of person-hours could be repurposed with this technology.

Impact of AI-Based Auto-Segmentation on Radiotherapy Processes in Low and Middle-Income Countries

Radiotherapy processes require significant human resources and expertise, creating a barrier for rapid deployment in low and middle-income countries (LMICs). Optimal radiotherapy (RT) relies on accurate segmentation of tumor targets and organs-at-risk (OARs) during the RT planning process. This study reports the impact of AI-based auto-segmentation on RT processes in an LMIC. Ten patients including five head and neck (HN), and five prostate cancer patients were randomly selected. Their planning CT images were subjected to auto- segmentation using an FDA-approved AI software tool, and manual segmentation by an experienced radiation oncologist from a Sub-Saharan African RT clinic. The control data consisted of contours from an experienced radiation oncologist and dosimetrists at a large academic institution in the US. For prostate cases, the contours included the prostate, seminal vesicles, bladder, rectum, penile bulb, and both femoral heads. For HN cases, the contours included the brain, brainstem, bilateral eyes, lens, optic nerves, cochlea, parotids, optic chiasm, spinal cord, oral cavity, and mandible. The time to complete the segmentation was recorded for both auto-segmentation and manual contours from the LMIC. The DICE similarity coefficients were used for comparative evaluation. The average time for contouring per patient was 2 minutes for AI compared to 57 minutes for manual contouring in the LMIC. When comparing the control data, AI pelvic contours provide a slightly better agreement than LMIC manual contours for all the OARs, with the following mean DICE coefficients for AI vs LMIC manual contours: bladder (0.971 vs 0.958), left femoral head (0.960 vs 0.949), right femoral head (0.959 vs 0.941), rectum (0.880 vs 0.867), prostate (0.836 vs 0.824), seminal vesicles (0.696 vs 0.580), and penile bulb (0.536 vs 0.528). For HN contours, AI provide a better agreement for 7 of 11 OARs than the LMIC manual contours, with the following mean DICE coefficients: brain (0.972 vs 0.982), mandible (0.877 vs 0.925), right parotid (0.847 vs 0.800), left parotid (0.798 vs 0.792), spinal cord (0.837 vs 0.821), left eye (0.875 vs 0.832), right eye (0.867 vs 0.836), brainstem (0.866 vs 852), oral cavity (0.796 vs 0.787), left lens (0.650 vs 0.729) and right lens (0.671 vs 0.682). Neither AI contours nor LMIC manual contours had good agreement with the control data (<0.600) for optic nerves, chiasm, and cochlea due to their small volumes. AI-based auto-segmentation tools are capable of producing contours of comparable quality to those generated by manual segmentation for both pelvic and HN cancer patients in LMICs, while also resulting in substantial time savings. AI-based auto-segmentation holds tremendous potential for improving radiotherapy care in LMICs with limited sources.

Anatomy-Guided Deep Learning Model for Accurate and Robust Gross Tumor Volume Segmentation in Lung Cancer Radiation Therapy

In lung cancer radiation therapy, clinicians must outline the gross tumor volume (GTV) precisely on the planning computed tomography (pCT) for accurate radiation dose delivery. However, due to the limited contrast between tumor and normal tissues in lung parenchyma, accurate delineation of tumor boundaries is difficult leading to large inter-observer variation. In this study, we develop an anatomy-guided lung GTV deep segmentation model using a training cohort of multi-center datasets. The quantitative segmentation performance is evaluated on an independent dataset, where the inter-observer delineation variation is also assessed. We collected and curated four publicly available lung datasets with GTV annotations (Lung-PET-CT-Dx, LIDC-IDRI, NSCLC-Radiogenomics and RIDER-CT) for deep learning model development. A total of 871 CT scans of patients, who were diagnosed with T1-T4 NSCLC, were available for training after data curation. The GTV annotations of primary tumor were examined and edited by two experienced radiation oncologists following the RTOG 1106 protocol. An anatomy-guided deep learning model was proposed, which consisted two deep networks. The first deep network used CT scan as input and segmented 4 anatomic organs (airway, heart, pulmonary artery and pulmonary vein), while the second deep network took both CT scan and these pre-segmented 4 organs as input and segmented the lung GTV. With the help of anatomic priors from 4 pre-segmented organs, the second deep network could more easily locate the GTV. We used nnUNet as the deep segmentation network. For evaluation, we used NSCLC-Radiomics as the testing dataset, which contains 20 CT scans each annotated by 5 radiation oncologists. The auto-segmented GTV were compared against each of the manual GTV reference. Inter-observer variation was also assessed using the 5 manual GTV references. The proposed anatomic-guided lung GTV segmentation model achieved a mean Dice score of 82.4% and 95% Hausdorff distance (HD95) of 6.9mm when averaged cross 20 patients and 5 GTV references (Table 1), which outperformed the basic deep GTV segmentation model by markedly reducing 19.4% HD95 error. The performance of proposed model was also comparable to the inter-observer variation (Dice score: 82.4% vs. 81.9%, HD95 6.9 vs. 6.4mm), indicating that our model had similar reproducibility as human observers. We developed and tested an anatomy-guided deep learning model for segmenting GTV in NSCLC patients. The model achieves high quantitative segmentation performance, which is comparable to the human observer variation. It can be potentially used in radiotherapy practice to improve GTV delineation consistency and reduce workloads of radiation oncologists.

Evaluation of a Deep Image to Image Network (DI2IN) Auto-Segmentation Algorithm across a Network of Cancer Centers

Due to manual OAR contouring challenges, various automatic contouring solutions have been introduced. Historically, common clinical auto-segmentation algorithms used were atlas based, which required maintaining a library of self-made contours. Searching the collection was computationally intensive and could take several minutes to complete. Deep learning approaches have shown significant benefits compared to atlas-based methods in improving segmentation accuracy and efficiency in auto-segmentation algorithms. This work represents the first multi-institutional study to describe and evaluate an AI algorithm for auto-segmentation of organs at risk (OARs) based on a deep image-to-image network (DI2IN). The AI-Rad Companion Organs RT (AIRC) algorithm (Siemens Healthineers, Erlangen, Germany) uses a two-step approach for segmentation. In the first step, the target organ region in the optimal input image is extracted using a trained Deep Reinforcement Learning network (DRL), which is then used as input to create the contours in the second step based on DI2IN. The study was initially designed as a prospective single-center evaluation. The automated contours generated by AIRC were evaluated by three experienced board-certified radiation oncologists using a 4-point scale where 4 is clinically usable and 1 requires re-contouring. After seeing favorable results in a single-center pilot study, we decided to expand the study to 6 additional institutions, encompassing 8 additional evaluators for a total of 11 physician evaluators across 7 institutions. One hundred fifty-six patients and 1366 contours were prospectively evaluated. The 5 most commonly contoured organs were the lung (136 contours, average rating = 4.0), spinal cord (106 contours, average rating = 3.1), eye globe (80 contours, average rating = 3.9), lens (77 contours, average rating = 3.9), and optic nerve (75 contours, average rating = 4.0). The average rating per evaluator per contour was 3.6. On average 124 contours were evaluated by each evaluator. 65% of the contours were rated as 4 and 31% were rated as 3. Only 4% of contours were rated as 1 or 2. 33 organs were evaluated in the study, with 19 structures having a 3.5 or above average rating (ribs, abdominopelvic cavity, skeleton, larynx, lung, aorta, brachial plexus, lens, eye globe, glottis, heart, parotid glands, bladder, kidneys, supraglottic larynx, submandibular glands, esophagus, optic nerve, oral cavity) and the remaining organs having a rating of 3.0 or greater (female breast, proximal femur, seminal vesicles, rectum, sternum, brainstem, prostate, brain, lips, mandible, liver, optic chiasm, spinal cord, spleen). No organ had an average rating below 3. AIRC performed well with greater than 95% of contours accepted by treating physicians with no or minor edits. It supported a fully automated workflow with the potential for time savings and increased standardization with the use of AI-powered algorithms for high quality OAR contouring.

Standardizing Breast Adjuvant Radiation Therapy Practices in a Diverse Health System through Development of a Delphi Consensus Supported Clinical Algorithm

Adjuvant radiation therapy (RT) for early breast cancer is a highly effective treatment option. However, the multitude of available techniques backed by strong clinical evidence can result in a lack of consistency in treatment approaches, even within a single healthcare organization. This presents challenges for both providers and patients in making informed decisions. To address this issue, our radiation department, which serves a large and diverse health system, developed a clinical algorithm for adjuvant RT for breast cancer. This algorithm was created to enhance the quality and standardization of care delivered across our network. A modified Delphi technique was used. A panel of eight experienced breast radiation oncologists from our institution was assembled. The panel first identified the common clinical scenarios encountered in treating patients with adjuvant breast RT, and then made recommendations for the primary and alternative approaches for each scenario. NCCN guidelines, ASTRO executive summaries, and published national and international randomized trials were used for reference. In case of disagreements, the final recommendation was reached through a majority vote. The draft algorithm was also shared with medical oncologists, surgeons, and patient advocates to gather their input prior to final approval by the expert panel. Consensus was reached for three broad clinical scenarios for patients who have undergone lumpectomy or mastectomy: Ductal Carcinoma in Situ (DCIS), Invasive Cancer Node Negative (ICN0), and Invasive Cancer Node Positive (ICN+). The panel agreed subdivision of the scenarios into three risk groups (low, intermediate, and high) and three age groups, based on guidelines and consensus statements. For DCIS patients, size, grade, margin status, hormone receptor status, and tumor focality were used for further stratification. For ICN0 patients, HER2 receptor status and lymphovascular space invasion were also included. For ICN+ patients, nodal status (negative versus 1-3 positive nodes versus more than 3 positive nodes) was used for stratification. Additionally, DCIS and ICN0 patients were further divided into age groups. The panel reached consensus recommendations for RT, including whole breast RT, partial breast RT, chest wall RT, regional nodal irradiation, or omission of RT for each sub-group. Clinical trial enrollment was also recommended where appropriate. A breast cancer adjuvant RT algorithm was developed with the aim of standardizing care for patients with breast cancer. Implementation is expected to standardize treatment recommendations in our health system and to streamline the shared decision-making process with patients.

Deep Learning-Based Multi-Modality Segmentation of Primary Gross Tumor Volume in CT and MRI for Nasopharyngeal Carcinoma

The delineation of primary gross tumor volume (GTV) of nasopharyngeal carcinoma (NPC) is an essential step for radiotherapy planning. In clinical practice, radiation oncologists manually delineate the GTV in planning CT with the help of diagnostic MRI. This is because NPC tumors are closely adjacent to many important anatomic structures, and CT and MRI provide complementary strength to accurately determine the tumor extension boundary. Manual delineation is time-consuming with the potential registration errors between MRI and CT decreasing the delineation accuracy. In this study, we propose a fully automated GTV segmentation method based on CT and MRI by first aligning MRI to CT, and then, segmenting the GTV using a multi-modality deep learning model. We collected 104 nasopharyngeal carcinoma patients with both planning CT and diagnostic MRI scans (T1 & T2 phases). An experienced radiation oncologists manually delineated the GTV, which was further examined by another senior radiation oncologist. Then, a coarse to fine cross-modality registration from MRI to CT was conducted as follows: (1) A rigid transformation was performed on MRI to roughly align MRI to CT with similar anatomic position. (2) Then, the region of interest (RoI) on both CT and rigid-transformed MRI were cropped. (3) A leading cross-modality deformable registration algorithm, named DEEDS, was applied on the cropped MRI and CT RoIs to find an accurate local alignment. Next, using CT and registered MRI as the combined input, a multi-modality deep segmentation network based on nnUNet was trained to generate the GTV prediction. 20% patients were randomly selected as the unseen testing set to quantitatively evaluate the performance. The quantitative NPC GTV segmentation performance is summarized in Table 1. The deep segmentation model using CT alone achieved reasonable high performance with 76.6% Dice score and 1.34mm average surface distance (ASD). When both CT and registered MRI were used, the segmentation model further improved the performance by 0.9% Dice score increase and 11% relative ASD error reduction, demonstrating the complementary strength of CT and MRI in determining NPC GTV. Notably, the achieved 77.5% Dice score and 1.19mm ASD by the multimodality model is among the top performing results reported in recent automatic NPC GTV segmentation using either CT or MRI modality. We developed a fully automated multi-modal deep-learning model for NPC GTV segmentation. The developed model can segment the NPC GTV in high accuracy. With further optimization and validation, this automated model has potential to standardize the NPC GTV segmentation and significantly decrease the workload of radiation oncologists in clinical practice.

Delineation of Lingual Artery as an Additional Organ-At-Risk for Stereotactic Body Radiation Therapy of Head and Neck Cancers

Lingual artery (LA) stenosis or bleeding can be a potentially fatal consequence of stereotactic body radiation therapy of recurrent head and neck cancer. We aim to provide delineation guidelines for LA contouring during head and neck radiation. Three experienced radiation oncologists and 1 radiologist studied the course of LA and its anatomic relationships. A step-by-step guideline was formulated for its delineation on contrast-enhanced axial computed tomography scans. We recommend that LA should be considered an important organ at risk for stereotactic body radiation therapy of head and neck cancer. Its delineation and dose constraints should be considered on a case-to-case basis.

A research on the improved rotational robustness for thoracic organ delineation by using joint learning of segmenting spatially-correlated organs: A U-net based comparison.

To study the improved rotational robustness by using joint learning of spatially-correlated organ segmentation (SCOS) for thoracic organ delineation. The network structure is not our point. The SCOS was implemented in a U-net-like model (abbr. SCOS-net) and evaluated on unseen rotated test sets. Two hundred sixty-seven patients with thoracic tumors (232 without rotation and 35 with rotation) were enrolled. The training and validation images came from 61 randomly chosen unrotated patients. The test data included two sets. One consisted of 3000 slices from the rest 171 unrotated patients. They were rotated by us by -30°∼30°. One was the images from the 35 rotated patients. The lung, heart, and spinal cord were delineated by experienced radiation oncologists and regarded as ground truth. The SCOS-net was compared with its single-task learning counterparts, two published multiple learning task settings, and rotation augmentation. Dice, 3 distance metrics (maximum and 95th percentile of Hausdorff distances and average surface distance (ASD)) and the number of cases where ASD = infinity were adopted. We analyzed the results using visualization techniques. In terms of no augmentation, the SCOS-net achieves the best lung and spinal cord segmentations and comparable heart delineation. With augmentation, SCOS performs better in some cases. The proposed SCOS can improve rotational robustness, and is promising in clinical applications for its low network capacity and computational cost.

A proposed framework for the implementation of head and neck cancer treatment at a new cancer center from a radiation oncology perspective.

Establishing a new head and neck cancer (HNC) treatment center requires multidisciplinary team management and expertise. To our knowledge, there are no clear recommendations or guidelines in the literature for the commencement of HNC radiation therapy (RT) at a new cancer center. We propose a novel framework outlining the necessary components required to set-up a new radiation therapy HNC treatment. We reviewed the infrastructure and methodology in the commencement of HNC radiation therapy in our cancer care center and invited several external, experienced metropolitan head and neck radiation oncologists to develop a novel consensus guideline that may be used by new RT centers to treat HNC. Recommendations were presented to our internal and external staff specialists using a survey questionnaire with ratings utilized to determine consensus using pre-defined thresholds as per the American Society of Clinical Oncology Guidelines Methodology Manual. This consensus recommendation aims to improve RT utilization whilst advocating for optimal patient outcomes by presenting a framework for new radiation therapy centers ready to step up and manage the treatment of head and neck cancer patients. We propose these evidence-based consensus guidelines endorsed by external HNC radiation oncologists.

Automated treatment planning for liver cancer stereotactic body radiotherapy.

To evaluate the quality of fully automated stereotactic body radiation therapy (SBRT) planning based on volumetric modulated arc therapy, which can reduce the reliance on historical plans and the experience of dosimetrists. Fully automated re-planning was performed on twenty liver cancer patients, automated plans based on automated SBRT planning (ASP) program and manual plans were conducted and compared. One patient was randomly selected and evaluate the repeatability of ASP, ten automated and ten manual SBRT plans were generated based on the same initial optimization objectives. Then, ten SBRT plans were generated for another selected randomly patient with different initial optimization objectives to assess the reproducibility. All plans were clinically evaluated in a double-blinded manner by five experienced radiation oncologists. Fully automated plans provided similar planning target volume dose coverage and statistically better organ at risk sparing compared to the manual plans. Notably, automated plans achieved significant dose reduction in spinal cord, stomach, kidney, duodenum, and colon, with a median dose of D2% reduction ranging from 0.64 to 2.85Gy. R50% and Dmean of ten rings for automated plans were significantly lower than those of manual plans. The average planning time for automated and manual plans was 59.8 ± 7.9min vs. 127.1 ± 16.8min (-67.3min). Automated planning for SBRT, without relying on historical data, can generate comparable or even better plan quality for liver cancer compared with manual planning, along with better reproducibility, and less clinically planning time.

Developing an AI-assisted planning pipeline for hippocampal avoidance whole brain radiotherapy

Background and purposeHippocampal avoidance whole brain radiotherapy (HA-WBRT) is effective for controlling disease and preserving neuro-cognitive function for brain metastases. However, contouring and planning of HA-WBRT is complex and time-consuming. We designed and evaluated a pipeline using deep learning tools for a fully automated treatment planning workflow to generate HA-WBRT radiotherapy plans. Materials and methodsWe retrospectively collected 50 adult patients who received HA-WBRT. Using RTOG- 0933 clinical trial protocol guidelines, all organs-at-risk (OARs) and the clinical target volume (CTV) were contoured by experienced radiation oncologists. A deep-learning segmentation model was designed and trained. Next, we developed a volumetric-modulated arc therapy (VMAT) auto-planning algorithm for 30 Gy in 10 fractions. Automated segmentations were evaluated using the Dice similarity coefficient (DSC) and 95th-percentile Hausdorff distance (95 % HD). Auto-plans were evaluated by the percentage of PTV volume that receives 30 Gy (V30Gy), conformity index (CI), and homogeneity index (HI) of planning target volume (PTV) and the minimum dose (D100%) and maximum dose (Dmax) for the hippocampus, Dmax for the lens, eyes, optic nerve, brain stem, and chiasm. ResultsWe developed a deep-learning segmentation model and an auto-planning script. For the 10 cases in the independent test set, the overall average DSC and 95 % HD of contours were greater than 0.8 and less than 7 mm, respectively. All auto-plans met the RTOG- 0933 criteria. The HA-WBRT plan automatically created time was about 10 min. ConclusionsAn artificial intelligence (AI)-assisted pipeline using deep learning tools can rapidly and accurately generate clinically acceptable HA-WBRT plans with minimal manual intervention and increase efficiency of this treatment for brain metastases.