Arterial spin labelling MRI predicts diagnostic label and progression on ADNI-3 data

Motivation: CBF is a surrogate biomarker for early AD detection and diagnostic progression, as it reflects vascular dysregulation linked to AD pathology. Goal(s): To predict AD diagnostic label and MCI conversion using ASL data from ADNI-3. Approach: We trained SVM classifiers with harmonised CBF values, demographic covariates and scanner types. Performance was validated across multiple binary classifications. Data augmentation and feature selection were used to enhance model robustness. Results: High specificity of 0.98 was achieved in distinguishing AD and CN/MCI groups, with limitations in CN vs. MCI. The classifiers also effectively identified stable vs. converting MCI cases. Impact: CBF-based classifiers using ASL-MRI offer a promising non-invasive approach to early AD diagnosis and tracking disease progression. It could add value to the current neuroimaging and fluid-based biomarkers.