Abstract
Background: Predicting the risk of disease progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD) has important clinical significance. This study aims at providing a personalized MCI-to-AD conversion prediction via radiomics-based predictive modeling (RPM) with multicentre 18F-Fluorodeoxyglucose positron emission tomography (FDG PET) data. Method: Three cohorts of 18F-FDG PET data and neuropsychological assessments were gathered from patients examined at Huashan Hospital (n=22), Xuanwu Hospital (n=80), and from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (n=355). Of these, amyloid images were selected for the ADNI and Xuanwu cohorts. First, 430 radiomic features were extracted from the 80 regions of interest (ROIs) for all PET images. These features were then concatenated for feature selection and an RPM model was constructed on the ADNI dataset. In addition, we used clinical scale data to establish a clinical Cox model, and a combined model for comparison. Afterwards, the images from Huashan Hospital were used to validate the stability and reliability of RPM, and the images from Xuanwu Hospital were used to explore the differences of biomarkers at different cognitive stages. Finally, correlation analysis was conducted between the radiomic biomarkers, neuropsychological assessments, and amyloid burden. Results: Experimental results show that the predictive performance of the PET-modal cox model was better than clinical cox model. In the two test data sets, the C index of PET model is 0.75 and 0.73, respectively; The C index of clinical model is 0.68. Moreover, most crucial image biomarkers had significant differences at different cognitive stages, and were significantly correlated with cognitive ability and the amyloid global level standardized uptake value ratio.Conclusion: The preliminary results demonstrated that the developed RPM approach has the potential to monitor the progress in high-risk populations with AD.
18F-FDG PET Database of Longitudinally Confirmed Healthy Elderly Individuals Improves Detection of Mild Cognitive Impairment and Alzheimer's Disease