Purpose: Radiation therapy treatment planning is a time‐consuming and iterative manual process. Consequently, plan quality varies greatly between and within institutions. Artificial intelligence shows great promise in improving plan quality and reducing planning times. This technical note describes our participation in the American Association of Physicists in Medicine Open Knowledge‐Based Planning Challenge (OpenKBP), a competition to accurately predict radiation therapy dose distributions.Methods: A three‐dimensional (3D) densely connected U‐Net with dilated convolutions was developed to predict 3D dose distributions given contoured CT images of head and neck patients as input. While traditional augmentation techniques such as rotations and translations were explored, it was found that training on random patches alone resulted in the greatest model performance. A custom‐weighted mean squared error loss function was employed. Finally, an ensemble of best‐performing networks was used to generate the final challenge predictions.Results: Our team (SuperPod) placed second in the dose stream of the OpenKBP challenge. The average mean absolute difference between the predicted and clinical dose distributions of the testing dataset was 2.56 Gy. On average, the predicted normalized target DVH metrics were within 3% of the clinical plans, and the predicted organ at risk DVH metrics were within 2 Gy of the clinical plans.Conclusions: The developed 3D dense dilated U‐Net architecture can accurately predict 3D radiotherapy dose distributions and can be used as part of a fully automated radiation therapy planning pipeline.

Radiomics is one such “big data” approach that applies advanced image refining/data characterization algorithms to generate imaging features that can quantitatively classify tumor phenotypes in a non-invasive manner. We hypothesize that certain textural features of oropharyngeal cancer (OPC) primary tumors will have statistically significant correlations to patient outcomes such as local control. Patients from an IRB-approved database dispositioned to (chemo)radiotherapy for locally advanced OPC were included in this retrospective series. Pretreatment contrast CT scans were extracted and radiomics-based analysis of gross tumor volume of the primary disease (GTVp) were performed using imaging biomarker explorer (IBEX) software that runs in Matlab platform. Data set was randomly divided into a training dataset and test and tuning holdback dataset. Machine learning methods were applied to yield a radiomic signature consisting of features with minimal overlap and maximum prognostic significance. The radiomic signature was adapted to discriminate patients, in concordance with other key clinical prognosticators. 465 patients were available for analysis. A signature composed of 2 radiomic features from pre-therapy imaging was derived, based on the Intensity Direct and Neighbor Intensity Difference methods. Analysis of resultant groupings showed robust discrimination of recurrence probability and Kaplan-Meier-estimated local control rate (LCR) differences between “favorable” and “unfavorable” clusters were noted.

To develop a deep learning model that generates consistent, high-quality lymph node clinical target volumes (CTV) contours for head and neck cancer (HNC) patients, as an integral part of a fully automated radiation treatment planning workflow. Computed tomography (CT) scans from 71 HNC patients were retrospectively collected and split into training (n = 51), cross-validation (n = 10), and test (n = 10) data sets. All had target volume delineations covering lymph node levels Ia through V (Ia-V), Ib through V (Ib-V), II through IV (II-IV), and retropharyngeal (RP) nodes, which were previously approved by a radiation oncologist specializing in HNC. Volumes of interest (VOIs) about nodal levels were automatically identified using computer vision techniques. The VOI (cropped CT image) and approved contours were used to train a U-Net autosegmentation model. Each lymph node level was trained independently, with model parameters optimized by assessing performance on the cross-validation data set. Once optimal model parameters were identified, overlap and distance metrics were calculated between ground truth and autosegmentations on the test set. Lastly, this final model was used on 32 additional patient scans (not included in original 71 cases) and autosegmentations visually rated by 3 radiation oncologists as being "clinically acceptable without requiring edits," "requiring minor edits," or "requiring major edits." When comparing ground truths to autosegmentations on the test data set, median Dice Similarity Coefficients were 0.90, 0.90, 0.89, and 0.81, and median mean surface distance values were 1.0 mm, 1.0 mm, 1.1 mm, and 1.3 mm for node levels Ia-V, Ib-V, II-IV, and RP nodes, respectively. Qualitative scoring varied among physicians. Overall, 99% of autosegmented target volumes were either scored as being clinically acceptable or requiring minor edits (ie, stylistic recommendations, <2 minutes). We developed a fully automated artificial intelligence approach to autodelineate nodal CTVs for patients with intact HNC. Most autosegmentations were found to be clinically acceptable after qualitative review when considering recommended stylistic edits. This promising work automatically delineates nodal CTVs in a robust and consistent manner; this approach can be implemented in ongoing efforts for fully automated radiation treatment planning.

Accurate clinical target volume (CTV) delineation is essential to ensure proper tumor coverage in radiation therapy. This is a particularly difficult task for head-and-neck cancer patients where detailed knowledge of the pathways of microscopic tumor spread is necessary. This paper proposes a solution to auto-segment these volumes in oropharyngeal cancer patients using a two-channel 3D U-Net architecture. The first channel feeds the network with the patient’s CT image providing anatomical context, whereas the second channel provides the network with tumor location and morphological information. Radiation therapy simulation computer tomography scans and their corresponding manually delineated CTV and gross tumor volume (GTV) delineations from 285 oropharyngeal patients previously treated at MD Anderson Cancer Center were used in this study. CTV and GTV delineations underwent rigorous group peer-review prior to the start of treatment delivery. The convolutional network’s parameters were fine-tuned using a training set of 210 patients using 3-fold cross-validation. During hyper-parameter selection, we use a score based on the overlap (dice similarity coefficient (DSC)) and missed volumes (false negative dice (FND)) to minimize any possible under-treatment. Three auto-delineated models were created to estimate tight, moderate, and wide CTV margin delineations. Predictions on our test set (75 patients) resulted in auto-delineations with high overlap and close surface distance agreement (DSC > 0.75 on 96% of cases for tight and moderate auto-delineation models and 97% of cases having mean surface distance ⩽ 5.0 mm) to the ground-truth. We found that applying a 5 mm uniform margin expansion to the auto-delineated CTVs would cover at least 90% of the physician CTV volumes for a large majority of patients; however, determination of appropriate margin expansions for auto-delineated CTVs merits further investigation.