Domain adversarial networks and intensity-based data augmentation for male pelvic organ segmentation in cone beam CT.

In radiation therapy, a CT image is used to manually delineate the organs and plan the treatment. During the treatment, a cone beam CT (CBCT) is often acquired to monitor the anatomical modifications. For this purpose, automatic organ segmentation on CBCT is a crucial step. However, manual segmentations on CBCT are scarce, and models trained with CT data do not generalize well to CBCT images. We investigate adversarial networks and intensity-based data augmentation, two strategies leveraging large databases of annotated CTs to train neural networks for segmentation on CBCT. Adversarial networks consist of a 3D U-Net segmenter and a domain classifier. The proposed framework is aimed at encouraging the learning of filters producing more accurate segmentations on CBCT. Intensity-based data augmentation consists in modifying the training CT images to reduce the gap between CT and CBCT distributions. The proposed adversarial networks reach DSCs of 0.787, 0.447, and 0.660 for the bladder, rectum, and prostate respectively, which is an improvement over the DSCs of 0.749, 0.179, and 0.629 for "source only" training. Our brightness-based data augmentation reaches DSCs of 0.837, 0.701, and 0.734, which outperforms the morphons registration algorithms for the bladder (0.813) and rectum (0.653), while performing similarly on the prostate (0.731). The proposed adversarial training framework can be used for any segmentation application where training and test distributions differ. Our intensity-based data augmentation can be used for CBCT segmentation to help achieve the prescribed dose on target and lower the dose delivered to healthy organs.

Impact of machine log-files uncertainties on the quality assurance of proton pencil beam scanning treatment delivery.

Irradiation log-files store useful information about the plan delivery, and together with independent Monte Carlo dose engine calculations can be used to reduce the time needed for patient-specific quality assurance (PSQA). Nonetheless, machine log-files carry an uncertainty associated to the measurement of the spot position and intensity that can influence the correct evaluation of the quality of the treatment delivery. This work addresses the problem of the inclusion of these uncertainties for the final verification of the treatment delivery. Dedicated measurements performed in an IBA Proteus Plus gantry with a pencil beam scanning (PBS) dedicated nozzle have been carried out to build a 'room-dependent' model of the spot position uncertainties. The model has been obtained through interpolation of the look-up tables describing the systematic and random uncertainties, and it has been tested for a clinical case of a brain cancer patient irradiated in a dry-run. The delivered dose has been compared with the planned dose with the inclusion of the errors obtained applying the model. Our results suggest that the accuracy of the treatment delivery is higher than the spot position uncertainties obtained from the log-file records. The comparison in terms of DVHs shows that the log-reconstructed dose is compatible with the planned dose within the 95% confidence interval obtained applying our model. The initial mean dose difference between the calculated dose to the patient based on the plan and recorded data is around 1%. The difference is essentially due to the log-file uncertainties and it can be removed with a correct treatment of these errors. In conclusion our new PSQA protocol allows for a fast verification of the dose delivered after every treatment fraction through the use of machine log-files and an independent Monte Carlo dose engine. Moreover, the inclusion of log-file uncertainties in the dose calculation allows for a correct evaluation of the quality of the treatment plan delivery.