Machine learning techniques based on 18F-FDG PET radiomics features of temporal regions for classification of temporal lobe epilepsy patients from healthy controls

Abstract

Abstract Background To investigate the clinical application of 18F-FDG PET radiomics features for temporal lobe epilepsy and create a radiomics-based model for differentiating TLE patients from healthy controls. Methods A total of 347 subjects that underwent 18F-FDG PET scans from March 2014 to January 2020 (234 TLE patients and 113 controls) were allocated to training (n = 242) and test (n = 105) sets. All PET images were registered to the Montreal Neurological Institute template. PyRadiomics was used to extract radiomics features from the temporal regions segmented according to the AAL atlas. The least absolute shrinkage and selection operator and Boruta algorithms were applied to select the radiomic features significantly associated with TLE. Eleven machine-learning algorithms were used to establish models. Results The final radiomics features (n = 22)used for model training were selected by the combinations of the least absolute shrinkage and selection operator and the Boruta algorithm with cross-validation. All data were randomly divided into a training set (n = 242) and a testing set (n = 105). Among eleven machine-learning algorithms, the logistic regression (AUC 0.984, F1-Score 0.959) performed the best of the 11 machine-learning models. The AUCs of the tuned logistic regression model in the training and test sets were 98.1 and 95.7. Conclusions The radiomics model from temporal regions can be a potential method for distinguishing TLE. Machine learning-based diagnosis of TLE from preoperative FDG PET images could serve as a useful preoperative diagnostic tool.