BackgroundDeep learning-based head and neck lymph node level (HN_LNL) autodelineation is of high relevance to radiotherapy research and clinical treatment planning but still underinvestigated in academic literature. In particular, there is no publicly available open-source solution for large-scale autosegmentation of HN_LNL in the research setting.MethodsAn expert-delineated cohort of 35 planning CTs was used for training of an nnU-net 3D-fullres/2D-ensemble model for autosegmentation of 20 different HN_LNL. A second cohort acquired at the same institution later in time served as the test set (n = 20). In a completely blinded evaluation, 3 clinical experts rated the quality of deep learning autosegmentations in a head-to-head comparison with expert-created contours. For a subgroup of 10 cases, intraobserver variability was compared to the average deep learning autosegmentation accuracy on the original and recontoured set of expert segmentations. A postprocessing step to adjust craniocaudal boundaries of level autosegmentations to the CT slice plane was introduced and the effect of autocontour consistency with CT slice plane orientation on geometric accuracy and expert rating was investigated.ResultsBlinded expert ratings for deep learning segmentations and expert-created contours were not significantly different. Deep learning segmentations with slice plane adjustment were rated numerically higher (mean, 81.0 vs. 79.6, p = 0.185) and deep learning segmentations without slice plane adjustment were rated numerically lower (77.2 vs. 79.6, p = 0.167) than manually drawn contours. In a head-to-head comparison, deep learning segmentations with CT slice plane adjustment were rated significantly better than deep learning contours without slice plane adjustment (81.0 vs. 77.2, p = 0.004). Geometric accuracy of deep learning segmentations was not different from intraobserver variability (mean Dice per level, 0.76 vs. 0.77, p = 0.307). Clinical significance of contour consistency with CT slice plane orientation was not represented by geometric accuracy metrics (volumetric Dice, 0.78 vs. 0.78, p = 0.703).ConclusionsWe show that a nnU-net 3D-fullres/2D-ensemble model can be used for highly accurate autodelineation of HN_LNL using only a limited training dataset that is ideally suited for large-scale standardized autodelineation of HN_LNL in the research setting. Geometric accuracy metrics are only an imperfect surrogate for blinded expert rating.

Purpose To share our experiences in implementing a dedicated magnetic resonance (MR) scanner for radiotherapy (RT) treatment planning using a novel coil setup for brain imaging in treatment position as well as to present developed core protocols with sequences specifically tuned for brain and prostate RT treatment planning. Materials and methods Our novel setup consists of two large 18-channel flexible coils and a specifically designed wooden mask holder mounted on a flat tabletop overlay, which allows patients to be measured in treatment position with mask immobilization. The signal-to-noise ratio (SNR) of this setup was compared to the vendor-provided flexible coil RT setup and the standard setup for diagnostic radiology. The occurrence of motion artifacts was quantified. To develop magnetic resonance imaging (MRI) protocols, we formulated site- and disease-specific clinical objectives. Results Our novel setup showed mean SNR of 163 ± 28 anteriorly, 104 ± 23 centrally, and 78 ± 14 posteriorly compared to 84 ± 8 and 102 ± 22 anteriorly, 68 ± 6 and 95 ± 20 centrally, and 56 ± 7 and 119 ± 23 posteriorly for the vendor-provided and diagnostic setup, respectively. All differences were significant (p > 0.05). Image quality of our novel setup was judged suitable for contouring by expert-based assessment. Motion artifacts were found in 8/60 patients in the diagnostic setup, whereas none were found for patients in the RT setup. Site-specific core protocols were designed to minimize distortions while optimizing tissue contrast and 3D resolution according to indication-specific objectives. Conclusion We present a novel setup for high-quality imaging in treatment position that allows use of several immobilization systems enabling MR-only workflows, which could reduce unnecessary dose and registration inaccuracies.