S658
Clinical – Head & neck
ESTRO 2026
For the same LNs auto-segmentation was performed. These autocontours were visually evaluated with a Lickert-scale of 1 to 5 (1 = rejection - 5 = no editing necessary). Subsequently, autocontours were edited (AE) until clinically acceptable. For both manual contouring and editing of autocontours the contouring-time was recorded and time saving calculated. To assess the interobserver-variability a CTVe, containing the clinically relevant LNs for each patient, was contoured manually and via editing of autocontours by two independent Radiation Oncologists (ROs). The similarity between manual contours and edited autocontours was expressed by Dice-Similarity-Coefficient (DSC) and surface-Dice- Similarity-Coefficient (s-DSC). Results: The manual contouring time for all LNs was median 39:08 min (range 33:26 – 62:24 min) per patient and this time was reduced to a median of 21:31 min (range 10:49 – 35:03 min) by using the auto-segmentation tool (45% time reduction). The reduction in contouring time (in minutes) from manual to edited autocontours per LN was as follows: Ia (0:34, 0:26), Ib (5:37, 2:45), II (8:38, 5:35), III (5:42, 2:33), IVa (3:25, 1:44), IVb (2:24, 2:16), V (4:28, 1:04), Vc (2:57, 1:19), VIa (3:14, 1:06), VIIa (2:12,1:46). None of the structures were rejected and the majority needed only minor editing, expressed as high values on the Lickert Scale (see Table 1). The median improvement in DSC and s-DSC from MC to AE for the CTVe for all patients between two ROs was 0.07 (0 - 0.11) and 0.13 (0.11 – 0.16). For the improvement in DSC and s-DSC, expressed as difference between AE and MC, for each patient (see Table 2).
ENE; major ENE significantly associated with higher pN status (p<0.01).Other factors significantly associated with pN status included PPOI type 4, WPOI type 5, DoI >1mm, LVI and PNI (all with p<0.01).DoI >1cm was associated with significantly worse DFS, increased recurrence and poorer local control (p<0.01, p=0.045).PNI presence led to significantly lower RFS and DFS (p=0.003, p=0.001).Cox proportional hazards analysis showed major ENE significantly associated with inferior RFS (HR=0.712, 95% CI: 0.599–0.847, p<0.001) and OS (HR=0.506, 95% CI: 0.381–0.671, p<0.001).Patients with DOI >1cm, involved margins (<1mm), PPOI type 4, WPOI type 5, LVI, and PNI who received surgery followed by chemoradiation had better local control. Conclusion: Major ENE independently predicts higher nodal stage and poorer local control, DFS, and OS, with additional adverse effects from PNI, LVI, WPOI, PPOI, and greater DOI. Patients with involved margins and major ENE derive maximal benefit from adjuvant chemoradiotherapy. Detailed histopathological risk grouping is crucial for optimizing postoperative treatment and predicting OCSCC survival outcomes. Keywords: Oral cavity SCC, ENE, Survival Analysis of contouring time and interobserver- variability for lymph node levels in head and neck cancer patients using deep-learning autosegmentation Christoph Dumke 1 , Steffen Barczyk 2 , Tobias Hofmann 1 , Norina Predescu 3 , Stefan Lautenschläger 4 1 Practice for Radiotherapy, Vivantes, Berlin, Germany. 2 Centre for Radiotherapy and radiation Oncology, Centre for Radiotherapy, Bocholt, Germany. 3 MVision Digital Poster 4395
AI, MVision AI, Helsinki, Finland. 4 Clinic for Radiotherapy, Clinic Lippe, Lemgo, Germany
Purpose/Objective: For extensively preoperated head and neck cancer patients (epHNCP) the delineation of lymph node levels (LN) can be challenging and therefore very time consuming. Especially for this situation, deep-learning- based auto-segmentation may be helpful. Therefore, we analyzed time saving and interobserver-variability between manually delineated and autocontoured LNs via an auto-segmentation tool (MVision Contour+™) in epHNCP. Material/Methods: Planning CT-scans of 10 epHNP with bilateral neck dissection and flap reconstruction were retrospectively obtained at the Vivantes Practice for Radiotherapy in Berlin-Spandau, Germany. Contouring of LNs (IA, IB, II, III, IVa, IVb, V, Vc, VIa, VIIa) was done manually (MC).
Made with FlippingBook - Share PDF online