USING SUBTRACTION MRI TO IMPROVE THE DETECTION OF AMYLOID-RELATED IMAGING ABNORMALITIES WITH EDEMA OR EFFUSION (ARIA-E) IN PATIENTS AFFECTED BY ALZHEIMER’S DISEASE RECEIVING IMMUNOTHERAPY: AN INTER-OBSERVER STUDY

Background: Amyloid-s plaques are considered the hallmark of Alzheimer's disease (AD). Immunotherapeutic strategies targeting Abeta have shown promising results, although they are associated with adverse events, including amyloid-related imaging abnormalities with edema and effusion (ARIA-E). A magnetic resonance imaging (MRI) scale to improve the detection and classification of these lesions has recently been approved. The purpose of this study was to investigate the use of the subtraction technique as an adjunct to standard FLAIR to improve the detection of ARIA-E using an established rating score. Methods:We included 75 AD patients who were submitted to Bapineuzumab treatment in a phase II study, 29 of whom developed ARIA-E. Five experienced neuroradiologists rated the brain MRI scans with and without using subtraction images after image registration. The accuracy of detection, intraclass correlation coefficient (ICC) and positive and negative specific agreement were calculated. Results: The sensitivity and specificity of ARIA-E cases detection per individual rater were excellent using both native images (1.00, 0.957, respectively) and subtraction images (0.966, 0.891, respectively). Overall, the detection rate with subtraction images resulted in higher sensitivity (0.966) and lower specificity rate (0.970) compared to native images (0.959, 0.991, respectively). ICC scores in both groups ranged from good to excellent, except for gyral swelling (moderate). Excellent negative specific agreement among all ARIA-E imaging features and good positive agreement were demonstrated in both groups. Positive agreement for subtraction images increased significantly when combining sulcal hyperintensity and gyral swelling. Conclusions: The use of subtraction imaging for the evaluation of ARIA-E may improve the detection of these abnormalities and the classification of their features. Nevertheless, this method would benefit from isotropic acquisition, enhanced training, and modification of the ARIA-E rating scale to lump parenchymal and sulcal hyperintensity or swelling. (Table Presented).