EUVIMEDEUVIMED entdecken
Crossref

Effect of ROI Definition on PET Image Features and Outcome Prediction in HNSCC

M Vácha, F Hofheinz, A Zwanenburg, F Lohaus, S Zschaeck, S Hoberück, P Nikulin, J Maus, E Lombardo, G Landry, A Holzgreve, M Krause und 3 weitere

Nuklearmedizin - NuclearMedicine · 01.04.2026 · 65 (02) · 132-133

DOI: 10.1055/s-0046-1818256 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Nuklearmedizin - NuclearMedicine, erschienen am 01.04.2026. Verfasst von M Vácha und 14 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ziel/Aim: Due to its rising incidence, head and neck squamous cell carcinoma (HNSCC) represents a global health challenge. In many cases, radio(chemo)therapy (RCHT) is an effective primary treatment, and PET-derived radiomic features – extracted from suitable Regions-Of-Interest (ROIs) – show considerable potential for predictions of its outcomes [ 1 ]. Two different primary tumor ROI definitions are commonly used: the Gross Tumor Volume (GTV) based on anatomical imaging and the semi-automated FDG-PET-based Metabolic Tumor Volume (MTV). In this study, we assessed the impact of the ROI definition on the extracted features and outcome predictions. Methodik/Methods: The study cohort comprised 739 patients with HNSCC who received primary RCHT at seven international Departments for Radiotherapy and Radiation Oncology. From both ROIs, GTV and MTV, and the corresponding FDG-PET image, standard radiomic features were extracted, and constant features were removed. Similarity metrics were measured for the ROIs (Dice similarity coefficient, DSC) and the resulting features (intraclass correlation coefficient, ICC). The associations between the features and RCHT outcomes (locoregional control, LRC; and overall survival, OS) were evaluated in univariate (UV) and multivariate (MV) analyses, adjusting for clinical covariates and PET image metadata. Ergebnisse/Results: Out of 171 evaluated features, only 22 were robust to the ROI definition (defined as ICC>0.90). The MTV ROIs yielded a higher number of features significantly associated with LRC and OS compared to the GTV ROIs, both in univariate and multivariate analyses (LRC: 42 vs. 22 in UV, 20 vs. 12 in MV; OS: 48 vs. 38 in UV, 24 vs. 13 in MV). Despite a strong overlap, some features (N=5) reached statistical significance only from the GTV ROI. Schlussfolgerungen/Conclusion: The findings indicate that the ROI definition (GTV vs. MTV) substantially affects radiomic feature values and their associations with clinical outcomes. Overall, MTV-based ROIs yielded more predictive features and appear more promising for the development of future biomarkers. Publication History Article published online: 02 April 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany

Zitieren

APA

Vácha, M., Hofheinz, F., Zwanenburg, A., Lohaus, F., Zschaeck, S., Hoberück, S., Nikulin, P., Maus, J., Lombardo, E., Landry, G., Holzgreve, A., Krause, M., Troost, E., Löck, S., & Braune, A. (2026). Effect of ROI Definition on PET Image Features and Outcome Prediction in HNSCC. Nuklearmedizin - NuclearMedicine, 65(02), 132-133. https://doi.org/10.1055/s-0046-1818256

Vancouver

Vácha M, Hofheinz F, Zwanenburg A, Lohaus F, Zschaeck S, Hoberück S, et al. Effect of ROI Definition on PET Image Features and Outcome Prediction in HNSCC. Nuklearmedizin - NuclearMedicine. 2026;65(02):132-133. doi:10.1055/s-0046-1818256

BibTeX
@article{Vcha2026,
  author = {Vácha, M and Hofheinz, F and Zwanenburg, A and Lohaus, F and Zschaeck, S and Hoberück, S and Nikulin, P and Maus, J and Lombardo, E and Landry, G and Holzgreve, A and Krause, M and Troost, E and Löck, S and Braune, A},
  title = {Effect of ROI Definition on PET Image Features and Outcome Prediction in HNSCC},
  journal = {Nuklearmedizin - NuclearMedicine},
  year = {2026},
  volume = {65},
  number = {02},
  pages = {132--133},
  doi = {10.1055/s-0046-1818256}
}
Kostenlos starten →