RESUMEN
Depending on the clinical situation, different combinations of lymph node (LN) levels define the elective LN target volume in head-and-neck cancer (HNC) radiotherapy. The accurate auto-contouring of individual LN levels could reduce the burden and variability of manual segmentation and be used regardless of the primary tumor location. We evaluated three deep learning approaches for the segmenting individual LN levels I−V, which were manually contoured on CT scans from 70 HNC patients. The networks were trained and evaluated using five-fold cross-validation and ensemble learning for 60 patients with (1) 3D patch-based UNets, (2) multi-view (MV) voxel classification networks and (3) sequential UNet+MV. The performances were evaluated using Dice similarity coefficients (DSC) for automated and manual segmentations for individual levels, and the planning target volumes were extrapolated from the combined levels I−V and II−IV, both for the cross-validation and for an independent test set of 10 patients. The median DSC were 0.80, 0.66 and 0.82 for UNet, MV and UNet+MV, respectively. Overall, UNet+MV significantly (p < 0.0001) outperformed other arrangements and yielded DSC = 0.87, 0.85, 0.86, 0.82, 0.77, 0.77 for the combined and individual level I−V structures, respectively. Both PTVs were also significantly (p < 0.0001) more accurate with UNet+MV, with DSC = 0.91 and 0.90, respectively. The accurate segmentation of individual LN levels I−V can be achieved using an ensemble of UNets. UNet+MV can further refine this result.
RESUMEN
Background: Proximal esophageal cancer (EC) is commonly treated with definitive chemoradiation (CRT). The radiation dose and type of chemotherapy backbone are still under debate. The objective of this study was to compare the treatment outcomes of contemporary CRT regimens.Material and Methods: In this retrospective observational cohort study, we included patients with locally advanced squamous cell cancer of the proximal esophagus, from 11 centers in the Netherlands, treated with definitive CRT between 2004 and 2014. Each center had a preferential CRT regimen, based on cisplatin (Cis) or carboplatin-paclitaxel (CP) combined with low (≤50.4 Gy) or high (>50.4 Gy) dose radiotherapy (RT). Differences in overall survival (OS) between CRT regimens were assessed using a fully adjusted Cox proportional hazards and propensity score (PS) weighted model. Safety profiles were compared using a multilevel logistic regression model.Results: Two hundred patients were included. Fifty-four, 39, 95, and 12 patients were treated with Cis-low-dose RT, Cis-high-dose RT, CP-low-dose RT, and CP-high-dose RT, respectively. Median follow-up was 62.6 months (95% CI: 47.9-77.2 months). Median OS (21.9 months; 95% CI: 16.9-27.0 months) was comparable between treatment groups (logrank p = .88), confirmed in the fully adjusted and PS weighted model (p > .05). Grades 3-5 acute adverse events were less frequent in patients treated with CP-low-dose RT versus Cis-high-dose RT (OR 3.78; 95% CI: 1.31-10.87; p = .01). The occurrence of grades 3-5 late toxicities was not different between treatment groups.Conclusion: Our study was unable to demonstrate a difference in OS between the CRT regimens, probably related to the relatively small sample size. Based on the superior safety profile, carboplatin and paclitaxel-based CRT regimens are preferred in patients with locally advanced proximal EC.
