F-FDOPA PET for the non-invasive prediction of glioma molecular parameters: a radiomics study.
Fiche publication
Date publication
mai 2021
Journal
Journal of nuclear medicine : official publication, Society of Nuclear Medicine
Auteurs
Membres identifiés du Cancéropôle Est :
Pr TAILLANDIER Luc, Pr GAUCHOTTE Guillaume, Pr VERGER Antoine, Dr RECH Fabien
Tous les auteurs :
Zaragori T, Oster J, Roch V, Hossu G, Chawki MB, Grignon R, Pouget C, Gauchotte G, Rech F, Blonski M, Taillandier L, Imbert L, Verger A
Lien Pubmed
Résumé
The assessment of gliomas by F-FDOPA PET imaging in adjunct to MRI showed high performance by combining static and dynamic features to non-invasively predict the isocitrate dehydrogenase (IDH) mutations and the 1p/19q co-deletion, which the World Health Organization classified as significant parameters in 2016. The current study evaluates whether other F-FDOPA PET radiomics features further improve performance and the contributions of each of these features to performance. Our study included seventy-two, retrospectively selected, newly diagnosed, glioma patients with F-FDOPA PET dynamic acquisitions. A set of 114 features, including conventional static features and dynamic features as well as other radiomics features were extracted and machine-learning models trained to predict IDH mutations and the 1p/19q co-deletion. Models were based on a machine-learning algorithm built from stable, relevant, and uncorrelated features selected by hierarchical clustering followed by a bootstrapped feature selection process. Models were assessed by comparing area under the curve (AUC) using a nested cross-validation approach. Feature importance was assessed using SHapley Additive exPlanations (SHAP) values. The best models were able to predict IDH mutations (logistic regression with L2 regularization) and the 1p/19q co-deletion (support vector machine with radial basis function kernel) with an AUC of 0.831[0.790;0.873] and 0.724[0.669;0.782] respectively. For the prediction of IDH mutations, dynamic features were the most important features in the model (TTP: 35.5%). In contrast, other radiomics features were the most useful for predicting the 1p/19q co-deletion (up to 14.5% of importance for the small zone low grey level emphasis) . F-FDOPA PET is an effective tool for the non-invasive prediction of glioma molecular parameters using a full set of amino-acid PET radiomics features. The contribution of each feature set shows the importance of systematically integrating dynamic acquisition for the prediction of the IDH mutations as well as developing the use of radiomics features in routine practice for the prediction the 1p/19q co-deletion.
Mots clés
18F-FDopa PET, Oncology: Brain, Research Methods, Statistical Analysis, WHO 2016 classification, glioma, machine learning, radiomics
Référence
J Nucl Med. 2021 May 20;: