Diagnosis of cognitive impairment in chronic kidney disease: A radiomics and machine learning approach with quantitative susceptibility mapping

BACKGROUND: Chronic kidney disease (CKD) alters magnetic susceptibility within the basal ganglia, contributing to cognitive impairment (CI). This study aims to develop a radiomics-based model using quantitative susceptibility mapping (QSM) and machine learning for diagnosing CKD-related CI. METHOD: A total of 161 CKD patients were prospectively recruited, with 113 in the training set and 48 in the test set. Radiomic features were extracted from basal ganglia nuclei on QSM images. After preprocessing and feature selection, multiple machine learning algorithms were evaluated. The final radiomics model was selected based on decision curve analysis (DCA) in the test cohort. A combined model was built by integrating clinical characteristics with the radiomics model using multivariable logistic regression. Model performance was assessed using receiver operating characteristic (ROC) analysis and DCA. RESULTS: DCA identified the putamen based support vector machine (SVM) radiomics model as the optimal model. It achieved AUCs of 0.929 (95 % CI 0.870-0.972) in the training set and 0.891 (95 % CI 0.786-0.972) in the test set. The combined model showed further improvement, yielding AUCs of 0.964 (95 % CI 0.928-0.989) and 0.933 (95 % CI 0.856-0.987). DCA indicated the highest net benefit for the combined model. CONCLUSION: QSM based radiomics of the putamen, especially when combined with clinical characteristics, may serve as a promising noninvasive approach for identifying CKD related CI.