Diffusion Radiomic Features of the Language Network Predict the Conversion from Mild Cognitive Impairment to Alzheimer's

BackgroundMild cognitive impairment (MCI) precedes Alzheimers disease (AD) in [~]40% of cases, with early language deficits distinguishing converters. This study develops a DTI radiomics model from language network gray matter to predict MCI to AD conversion and identify preclinical biomarkers. MethodsThis retrospective case-control study analyzed diffusion tensor imaging (DTI) data from 97 individuals with MCI (29 converters, 68 non-converters) from the Alzheimers Disease Neuroimaging Initiative (ADNI). Ethical approval and participant consent were obtained by ADNI. Radiomic features were extracted from fractional anisotropy (FA) and mean diffusivity (MD) maps within language network gray matter. A logistic regression model using eleven selected features performed classification. Performance was evaluated using area under the receiver operating characteristic curve (AUC). Radiomic-cognitive associations were analyzed using Pearson correlations; group differences were assessed with Fishers r-to-z transformation. ResultsThe model achieved cross-validation AUC = 0.84 and test AUC = 0.83. SHAP analysis identified two top predictors: lower right temporal pole original_glcm_Correlation_FA and higher right frontal orbital cortex original_glszm_SmallAreaHighGrayLevelEmphasis_FA. Right frontal orbital cortex original_glszm_SmallAreaHighGrayLevelEmphasis_FA correlated positively with ADAS-Q4 in non-converters (r = 0.27, p < 0.001) but negatively in converters (r = -0.48, p < 0.001). ConclusionsA DTI radiomics model achieved AUC = 0.83 for predicting MCI to AD conversion, with bilateral language network microstructural features showing group-specific cognitive associations, supporting their potential as early Alzheimers risk biomarkers. Key pointsO_LINew method identifies Alzheimers risk before significant cognitive decline occurs C_LIO_LIBrain language regions show detectable changes in future converters C_LIO_LITexture analysis reveals early disease signatures in brain tissue C_LI