Retrospective assessment of patterns of recurrence relative to radiotherapy volumes for adjuvant conformal chemoradiotherapy in gastric cancer.

BACKGROUND: Since the Intergroup 0116 study was published in 2000, adjuvant postoperative chemoradiotherapy using CT-planned and 3D conformal/intensity-modulated radiotherapy has been offered routinely to fit patients with resected gastric cancer at Princess Margaret Hospital .The objective of this study was to analyze patterns of disease recurrence with respect to the radiotherapy volumes. METHODS: For the date and site (local, locoregional, or distant) of the first recurrence, medical records were reviewed for all patients treated at Princess Margaret Hospital with adjuvant chemoradiotherapy for resected gastric adenocarcinoma (January 1, 2000 to November 30, 2009). Patients whose recurrences were limited to local and/or regional sites were selected for further analysis. Available diagnostic imaging of the recurrence site was registered to the original planning radiotherapy dataset for contouring. If necessary to respect changes in anatomy, the contour was translocated on the basis of anatomic descriptors. The center of mass for each recurrence was identified as a point and its location was categorized according to the isodose encompassing it; in field (90 % or more), marginal (50-89 %), or out of field (less than 50 %). RESULTS: Of all 197 patients, 14 (7 %) had isolated locoregional failure, constituting 20 % of all 71 patients with a recurrence. Successful fusions were feasible in five cases. Of these recurrences, four were in field and one was marginal. In a further four cases, visual inspection was used, showing one in-field recurrence, one marginal recurrence, and two out-of-field recurrences. In five patients, either a useable original dataset or diagnostic imaging of the recurrence was not available. CONCLUSIONS: The rates of isolated local/locoregional tumor recurrence in this study were low. Of the small number of recurrences available for analysis, most (five of nine) were in field. Further studies involving a larger cohort of patients might allow a more meaningful analysis of trends in the recurrence site with evolving radiotherapy techniques.

Automatic classification of dental artifact status for efficient image veracity checks: effects of image resolution and convolutional neural network depth.

Enabling automated pipelines, image analysis and big data methodology in cancer clinics requires thorough understanding of the data. Automated quality assurance steps could improve the efficiency and robustness of these methods by verifying possible data biases. In particular, in head and neck (H&N) computed-tomography (CT) images, dental artifacts (DA) obscure visualization of structures and the accuracy of Hounsfield units; a challenge for image analysis tasks, including radiomics, where poor image quality can lead to systemic biases. In this work we analyze the performance of three-dimensional convolutional neural networks (CNN) trained to classify DA statuses. 1538 patient images were scored by a single observer as DA positive or negative. Stratified five-fold cross validation was performed to train and test CNNs using various isotropic resampling grids (643, 1283 and 2563), with CNN depths designed to produce 323, 163, and 83 machine generated features. These parameters were selected to determine if more computationally efficient CNNs could be utilized to achieve the same performance. The area under the precision recall curve (PR-AUC) was used to assess CNN performance. The highest PR-AUC (0.92 ± 0.03) was achieved with a CNN depth = 5, resampling grid = 256. The CNN performance with 2563 resampling grid size is not significantly better than 643 and 1283 after 20 epochs, which had PR-AUC = 0.89 ± 0.03 (p -value = 0.28) and 0.91 ± 0.02 (p -value = 0.93) at depths of 3 and 4, respectively. Our experiments demonstrate the potential to automate specific quality assurance tasks required for unbiased and robust automated pipeline and image analysis research. Additionally, we determined that there is an opportunity to simplify CNNs with smaller resampling grids to make the process more amenable to very large datasets that will be available in the future.

External validation and transfer learning of convolutional neural networks for computed tomography dental artifact classification.

Quality assurance of data prior to use in automated pipelines and image analysis would assist in safeguarding against biases and incorrect interpretation of results. Automation of quality assurance steps would further improve robustness and efficiency of these methods, motivating widespread adoption of techniques. Previous work by our group demonstrated the ability of convolutional neural networks (CNN) to efficiently classify head and neck (H&N) computed-tomography (CT) images for the presence of dental artifacts (DA) that obscure visualization of structures and the accuracy of Hounsfield units. In this work we demonstrate the generalizability of our previous methodology by validating CNNs on six external datasets, and the potential benefits of transfer learning with fine-tuning on CNN performance. 2112 H&N CT images from seven institutions were scored as DA positive or negative. 1538 images from a single institution were used to train three CNNs with resampling grid sizes of 643, 1283 and 2563. The remaining six external datasets were used in five-fold cross-validation with a data split of 20% training/fine-tuning and 80% validation. The three pre-trained models were each validated using the five-folds of the six external datasets. The pre-trained models also underwent transfer learning with fine-tuning using the 20% training/fine-tuning data, and validated using the corresponding validation datasets. The highest micro-averaged AUC for our pre-trained models across all external datasets occurred with a resampling grid of 2563 (AUC = 0.91 ± 0.01). Transfer learning with fine-tuning improved generalizability when utilizing a resampling grid of 2563 to a micro-averaged AUC of 0.92 ± 0.01. Despite these promising results, transfer learning did not improve AUC when utilizing small resampling grids or small datasets. Our work demonstrates the potential of our previously developed automated quality assurance methods to generalize to external datasets. Additionally, we showed that transfer learning with fine-tuning using small portions of external datasets can be used to fine-tune models for improved performance when large variations in images are present.

Quality assurance for the geometric accuracy of cone-beam CT guidance in radiation therapy.

The introduction of volumetric X-ray image-guided radiotherapy systems allows improved management of geometric variations in patient setup and internal organ motion. As these systems become a routine clinical modality, we propose a daily quality assurance (QA) program for cone-beam computed tomography (CBCT) integrated with a linear accelerator. The image-guided system used in this work combines a linear accelerator with conventional X-ray tube and an amorphous silicon flat-panel detector mounted orthogonally from the accelerator central beam axis. This article focuses on daily QA protocols germane to geometric accuracy of the CBCT systems and proposes tolerance levels on the basis of more than 3 years of experience with seven CBCT systems used in our clinic. Monthly geometric calibration tests demonstrate the long-term stability of the flex movements, which are reproducible within +/-0.5 mm (95% confidence interval). The daily QA procedure demonstrates that, for rigid phantoms, the accuracy of the image-guided process can be within 1 mm on average, with a 99% confidence interval of +/-2 mm.

Practical considerations arising from the implementation of lung stereotactic body radiation therapy (SBRT) at a comprehensive cancer center.

INTRODUCTION: With the anticipation of improved outcomes, especially for patients with early-stage non-small cell lung cancer, stereotactic body radiation therapy (SBRT) has been rapidly introduced into the thoracic radiation oncology community. Although at first glance lung SBRT might seem methodologically similar to conventional radiotherapy, there are important differences in its execution that require particular consideration. The objective of this paper is to highlight these and other issues to contribute to the safe and effective diffusion of lung SBRT. We discuss practical challenges that have been encountered in the implementation of lung SBRT at a single, large institution and emphasize the importance of a systematic approach to the design of lung SBRT services. METHODS: Specific technical and clinical components that were identified as being important during the development of lung SBRT at Princess Margaret Hospital are described. The clinical system that evolved from these is outlined. RESULTS: Using this clinical framework the practical topics addressed include: patient assessment, simulation and treatment planning, tumor and organ at risk delineation, trial set up before treatment, on-line image-guidance, and patient follow-up. CONCLUSIONS: The potential gain in therapeutic ratio that is theoretically possible with lung SBRT can only be realized if the tumor is adequately irradiated and normal tissue spared. A discussion of the component parts of lung SBRT is presented. It is a complex process and specific challenges need to be overcome to effect the satisfactory transition of lung SBRT into routine practice.