Simple delineations cannot substitute full 3d tumor delineations for MR-based radiomics prediction of locoregional control in oropharyngeal cancer

• Published in: European journal of radiology 2022-03, Vol.148, p.110167-110167, Article 110167
• Main Authors: Bos, Paula, van den Brekel, Michiel W.M., Taghavi, Marjaneh, Gouw, Zeno A.R., Al-Mamgani, Abrahim, Waktola, Selam, Aerts, Hugo J.W.L., Beets-Tan, Regina G.H., Castelijns, Jonas A., Jasperse, Bas
• Format: Article
• Language: English
• Subjects: Head and neck neoplasms; Head and Neck Neoplasms - diagnostic imaging; Humans; Machine learning; Magnetic Resonance Imaging; Oropharyngeal neoplasms; Oropharyngeal Neoplasms - diagnostic imaging; Outcome prediction; Radiomics; Retrospective Studies; Squamous Cell Carcinoma of Head and Neck
• ISSN: 0720-048X; 1872-7727
• DOI: 10.1016/j.ejrad.2022.110167

Manual delineation of head and neck tumor contours for radiomics analyses is tedious and time consuming. This study investigates if fast or readily available tumor contours can substitute full tumor contours by an experienced observer for an MR-based radiomics model to predict locoregional control (LRC) in oropharyngeal squamous cell carcinoma (OPSCC) tumors. Radiomic features were extracted from postcontrast T1-weighted MRIs of 177 OPSCC primary tumors using six different manual delineation strategies. LRC prediction models based on recursive feature elimination combined with logistic regression were built. Models were trained and tested on data from each separate delineation. Additionally, the model derived from segmentations from the experienced reader was tested by each of the alternative delineations. Complementary, this was repeated with removal of size and shape features. Model performance was evaluated using area under the curve (AUC). Prediction performance of the experienced radiologist tumor delineation (AUC: 0.74) was superior compared to all other delineations when trained and tested (AUCs: 0.41–0.56) or trained on experienced delineations and tested (AUC: 0.56–0.67) on alternative segmentations. Removal of size and shape features considerably decreases prediction performance (AUC: 0.54). Applying the model based on expert delineations to spherical or single slice delineations makes prediction worthless since these models predict one class. Fast or readily available contours cannot substitute full expert tumor delineations in radiomics models predictive of LRC in OPSCC.