Decoding Choroid Plexus Pathology in Alzheimer's Disease: A Longitudinal Radiomics Approach for Prodromal Identification and Risk Stratification.

AIMS: Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder, and its identification of its prodromal stages remains a challenge. The choroid plexus (CP) plays a crucial role in AD pathology. This study aims to develop the CP radiomics model to distinguish AD from mild cognitive impairment (MCI) patients and predict the risk of MCI progression to AD.

METHODS

Radiomics features of CP were derived from magnetic resonance imaging (MRI) data, utilizing scans derived from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort and a local institutional cohort. A range of 12 classic machine learning algorithms was utilized for model construction. Additionally, we assessed the partial correlations of CP radiomics features with the Mini-Mental State Examination (MMSE) score and with CSF biomarkers.

RESULTS

In the MCI versus AD classification task, the CP model attained a best AUC of 0.794. This performance was further boosted to an AUC of 0.907 upon integration with clinical features. In the model predicting MCI-to-AD conversion, the CP model achieved an AUC of 0.745, which increased to 0.908 after incorporating clinical features. Furthermore, the high-risk group, as defined by the radiomics model, had a significantly shorter time to AD conversion (HR = 2.201, p < 0.001). CP radiomics features showed significant correlations with MMSE and CSF biomarkers (p < 0.05).

CONCLUSION

The machine learning model based on CP radiomics features effectively differentiates AD from MCI patients and predicts the risk of MCI progression to AD, offering new insights into the role of the CP in AD pathophysiology.