Patterns of local residual disease and local failure after intensity modulated/image guided radiation therapy for sinonasal tumors in dogs

Automatic gross tumor segmentation of canine head and neck cancer using deep learning and cross-species transfer learning

Background

Radiotherapy (RT) is increasingly being used on dogs with spontaneous head and neck cancer (HNC), which account for a large percentage of veterinary patients treated with RT. Accurate definition of the gross tumor volume (GTV) is a vital part of RT planning, ensuring adequate dose coverage of the tumor while limiting the radiation dose to surrounding tissues. Currently the GTV is contoured manually in medical images, which is a time-consuming and challenging task.

Purpose

The purpose of this study was to evaluate the applicability of deep learning-based automatic segmentation of the GTV in canine patients with HNC.

Materials and methods

Contrast-enhanced computed tomography (CT) images and corresponding manual GTV contours of 36 canine HNC patients and 197 human HNC patients were included. A 3D U-Net convolutional neural network (CNN) was trained to automatically segment the GTV in canine patients using two main approaches: (i) training models from scratch based solely on canine CT images, and (ii) using cross-species transfer learning where models were pretrained on CT images of human patients and then fine-tuned on CT images of canine patients. For the canine patients, automatic segmentations were assessed using the Dice similarity coefficient (Dice), the positive predictive value, the true positive rate, and surface distance metrics, calculated from a four-fold cross-validation strategy where each fold was used as a validation set and test set once in independent model runs.

Results

CNN models trained from scratch on canine data or by using transfer learning obtained mean test set Dice scores of 0.55 and 0.52, respectively, indicating acceptable auto-segmentations, similar to the mean Dice performances reported for CT-based automatic segmentation in human HNC studies. Automatic segmentation of nasal cavity tumors appeared particularly promising, resulting in mean test set Dice scores of 0.69 for both approaches.

Conclusion

In conclusion, deep learning-based automatic segmentation of the GTV using CNN models based on canine data only or a cross-species transfer learning approach shows promise for future application in RT of canine HNC patients.

Relative tumor volume has prognostic relevance in canine sinonasal tumors treated with radiation therapy: A retrospective study

Tumor volume is controversially discussed as a prognostic factor in dogs treated with radiation therapy for sinonasal tumors. Dogs’ body sizes vary widely and relative rather than absolute tumor volume might provide better prognostic information. Our hypothesis was that relative rather than absolute tumor volume (gross tumor volume, GTV) influences time to progression (TTP) and that a larger tumor volume is correlated with a higher tumor stage. We retrospectively investigated possible correlations of initial GTV to weight, body surface area (BSA), nasal cavity size and the tumor stage in 49 dogs with sinonasal tumors. Here, also presumed sinonasal tumors, esthesioneuroblastomas and histologically benign tumors were included. The possible impact of absolute and relative GTV on response and outcome were assessed according to imaging findings in 34 dogs with available follow-up computed tomographies (CTs) after definitive-intent radiation therapy with either a regular (10x4.2 Gy) or a simultaneously- integrated boost protocol (SIB; GTV boosted to 10x4.83 Gy). In contrast to absolute GTV (p<0.001), the relative GTVs were not correlated with dogs’ body sizes. Absolute GTV, GTV relative to weight and BSA were not associated with TTP based on CT imaging. However, GTV relative to nasal cavity showed a prognostic influence with a hazard ratio of 10.97 (95%CI:1.25–96.06). When looking at GTV relative to nasal cavity, stage 3 and 4 tumors were significantly larger than stage 1 and 2 tumors (p = 0.005). Our results suggest that GTV relative to nasal cavity could be prognostic for TTP and a larger tumor volume relative to nasal cavity is correlated with a higher tumor stage.

Employing deep convolutional neural networks for segmenting the medial retropharyngeal lymph nodes in CT studies of dogs

While still in its infancy, the application of deep convolutional neural networks in veterinary diagnostic imaging is a rapidly growing field. The preferred deep learning architecture to be employed is convolutional neural networks, as these provide the structure preferably used for the analysis of medical images. With this retrospective exploratory study, the applicability of such networks for the task of delineating certain organs with respect to their surrounding tissues was tested. More precisely, a deep convolutional neural network was trained to segment medial retropharyngeal lymph nodes in a study dataset consisting of CT scans of canine heads. With a limited dataset of 40 patients, the network in conjunction with image augmentation techniques achieved an intersection-overunion of overall fair performance (median 39%, 25 percentiles at 22%, 75 percentiles at 51%). The results indicate that these architectures can indeed be trained to segment anatomic structures in anatomically complicated and breed-related variating areas such as the head, possibly even using just small training sets. As these conditions are quite common in veterinary medical imaging, all routines were published as an open-source Python package with the hope of simplifying future research projects in the community.

Patterns of local residual disease and local failure after intensity modulated/image guided radiation therapy for sinonasal tumors in dogs

BACKGROUND

Most dogs with sinonasal tumors (SNT) treated with radiation therapy (RT) died because of local disease progression.

HYPOTHESIS/OBJECTIVES

Our hypothesis is that the majority of local failure and residual disease would occur within the radiation field.

ANIMALS

Twenty-two dogs with SNT treated with RT.

METHODS

Retrospective cohort study.

INCLUSION CRITERIA

dogs with SNT receiving 10 daily fractions of 4.2 Gy with intensity modulated radiation therapy (IMRT)/image guided radiation therapy (IGRT) and follow-up cone beam computed tomography (CBCT). Each CBCT was registered with the original radiation planning CT and the gross tumor volume (GTV) contoured. The GTV was classified as residual (GTVr) or a failure (GTVf). The dose statistic for each GTV was calculated with the original IMRT plan. For GTVf, failures were classified as "in-field," "marginal," or "out-field" if at least 95, 20-95, or less than 20% of the volume of failure was within 95% (D95) of the total prescription dose, respectively.

RESULTS

There were 52 follow-up CBCT/CTs. Overall there was a GTVr for 20 dogs and GTVf for 16 dogs. The majority of GTVr volume was within the original GTV. GTVf analysis showed that 75% (12/16) were "in-field," 19% (3/16) were "marginal" and 6% (1/16) were "out-field."

CONCLUSION AND CLINICAL IMPORTANCE

In-field failures are the main pattern for local recurrence, and there is evidence of radioresistant subvolumes within the GTV.