Predicting Amyloid Pathology in Mild Cognitive Impairment Using Radiomics Analysis of Magnetic Resonance Imaging

Background: Noninvasive identification of amyloid-? (A?) is important for better clinical management mild cognitive impairment (MCI) patients. Objective: To investigate whether radiomics features in the hippocampus MCI improve prediction cerebrospinal fluid (CSF) A?42 status when integrated with profiles. Methods: A total 407 subjects from Alzheimer’s Disease Neuroimaging Initiative were allocated to training (n = 324) and test 83) sets. Radiomics 214) bilateral extracted magnetic resonance imaging (MRI). cut-off <192 pg/mL was applied define CSF status. After feature selection, random forest subsampling methods utilized develop three models which predict A?42: 1) a model; 2) model based on profiles; 3) combined The performances thereof validated set. using volume also developed validated. Results: best-performing showed an area under curve (AUC) 0.674 AUC 0.758 0.823 hippocampal lower performance, 0.543 Conclusion: MRI can help patients potentially triage invasive costly A? tests.