Free water in the hippocampal cingulum as a Radiomic biomarker for Identifying inflammatory neuropsychiatric Lupus: A cross-sectional case-control study

PURPOSE: Neuropsychiatric systemic lupus erythematosus (NPSLE) is a serious manifestation of systemic lupus erythematosus (SLE), yet its neuroimaging diagnosis remains challenging. This study aims to explore the value of FW-corrected diffusion model parameters in characterizing white matter microstructural alterations in patients with NPSLE. MATERIALS AND METHODS: This cross-sectional, case-control study enrolled patients with 33 inflammatory NPSLE patients, 24 ischemic NPSLE patients, and 33 SLE patients without neuropsychiatric manifestations (non-NPSLE) at the time of inclusion between September 2023 and March 2024. All participants underwent diffusion MRI. Tract-based spatial statistics (TBSS) were used to compare white matter differences among groups. Regions showing significant differences were used as regions of interest (ROI) to extract mean FW values. These imaging features were combined with clinical indicators (age, Montreal Cognitive Assessment (MoCA) scores, Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2K) scores) to construct multiple machine learning classification models. Model performance was evaluated using internal and external validation. RESULTS: TBSS analysis revealed that only the FW parameter showed significant differences between inflammatory NPSLE and non-NPSLE patients, specifically manifested as significantly elevated FW values in the bilateral hippocampal cingulum bundles. In internal validation, Linear Discriminant Analysis (LDA), Adaptive Boosting (AdaBoost), and Logistic Regression (LR) classifiers demonstrated optimal performance (area under the curve (AUC) = 0.910, accuracy 85.0%). In external validation, LDA and LR classifiers achieved the highest AUC value (0.956) and accuracy (92.86%). CONCLUSION: Elevated FW in the bilateral hippocampal cingulum bundles of inflammatory NPSLE patients likely indicates neuroinflammation. A diagnostic model combining FW parameters from this region with clinical indicators shows strong potential for distinguishing inflammatory from non-inflammatory NPSLE.