Radiomics features of FDG PET images predict ApoE4

FDG PET can give early clues for diagnosing Alzheimer’s disease. Our objective was to use the same scan to predict ApoE4, known to be a risk factor. Additionally, we aimed to identify brain regions and imaging features associated with ApoE4 to enhance understanding of disease mechanisms. We analyzed data from 112 participants from the ADNI database. PET and sMRI scans were co-registered using ANTs, with Freesurfer delineating 95 whole-brain, 19 hippocampal, and 9 amygdala regions. PyRadiomics extracted 120 features from these segments. Three feature selection techniques (ANOVA, PCA, and LASSO) were applied to the preprocessed features. The performance of these feature selection methods with nine classifiers AB, DT, GNB, GP, GB, HGB, MLP, QDA, and RF was evaluated using stratified k-fold cross-validation. Based on AUC and accuracy, the best combination of feature selection and classifier was selected. The extracted features were then further analyzed and reduced for interpretability. The proposed radiomics model predicted the homozygous ApoE4 genotype, with an AUC of 0.945 an accuracy of 0.875, a sensitivity of 0.914, and a specificity of 0.838. Reducing the feature set to seven key features resulted in a slightly decreased performance (AUC: 0.889). The hippocampus, entorhinal cortex, amygdala, thalamus, and pars orbitalis exhibited significant associations with this genotype. FDG PET can be used to predict the presence of the ApoE4 genotype. The key features driving model predictions may provide potential insights into the underlying pathophysiology and potentially lead to a more comprehensive interpretation of FDG PET studies. • This study pioneers FDG PET data analysis for ApoE4 genotype prediction via radiomics. • Key ROIs—hippocampus, entorhinal cortex, amygdala, thalamus, pars orbitalis were identified as crucial in ApoE4 detection. • Improved Clinical Practice: FDG PET imaging may help early identification of high-risk Alzheimer's individuals. • Field Contribution: Advances neuroimaging and Alzheimer's research with a new method to predict ApoE4 genotype.