Lecanemab in practice: AI-derived MRI predictors of benefit and Amyloid Related Imaging Abnormalities (ARIA).
Authors
Affiliations (8)
Affiliations (8)
- Cognitive Neurology Unit, Neurological Institute, Tel Aviv Sourasky Medical Center, 6 Weizmann St, Tel Aviv, 6423906, Israel. [email protected].
- Department of Neurology and Neurosurgery, Gray Faculty of Medical & Health Sciences, Tel Aviv University, Tel Aviv, Israel. [email protected].
- Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel. [email protected].
- icometrix, Leuven, Belgium.
- Cognitive Neurology Unit, Neurological Institute, Tel Aviv Sourasky Medical Center, 6 Weizmann St, Tel Aviv, 6423906, Israel.
- Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel.
- Department of Radiology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel.
- Department of Neurology and Neurosurgery, Gray Faculty of Medical & Health Sciences, Tel Aviv University, Tel Aviv, Israel.
Abstract
Lecanemab, a monoclonal antibody targeting amyloid beta, has demonstrated meaningful clinical benefits in early Alzheimer's disease (AD), yet real-world data is needed to optimize patient selection and enhance safety monitoring, particularly with respect to amyloid-related imaging abnormalities (ARIA). Integration of quantitative and AI-derived MRI biomarkers may improve risk stratification and prediction of clinical trajectory. We conducted a retrospective real-world study of eighty-two patients with biomarker-confirmed early AD who initiated lecanemab at Tel Aviv Sourasky Medical Center between November 2023 and June 2025. Baseline MRI included volumetric T1-weighted imaging and susceptibility-weighted imaging (SWI). Automated whole-brain, regional cortical, and hippocampal volumes, and percentiles were extracted using FDA-cleared AI tools (icobrain by icometrix). Microhaemorrhage (MH) burden was assessed by both human and AI-assisted reads. Cognitive outcomes were evaluated using change in Mini-Mental State Examination (MMSE). Linear regression models assessed MRI predictors of cognitive response, and multivariable logistic regression identified predictors of ARIA. Patients exhibited significantly lower cerebral volumes at treatment initiation. Mean whole brain percentile, mean gray-matter (GM) percentile, and mean white matter percentile were 11.45%, 8.6% and 38% respectively. Higher baseline GM volume predicted less MMSE decline at 12 months (β = 0.64, FDR-corrected p < 0.003). Hippocampal and white-matter volumes were not associated with cognitive outcomes. Seventeen patients (20.7%) developed ARIA. Baseline MH burden was the strongest predictor of ARIA (human rated OR=3.48 per MH, p=0.015, icobrain rated OR=3.25, p=0.01), while APOE ε4 carriage showed a strong directional trend which did not reach significance. Aspirin use and hypertension were not associated with ARIA. Agreement between icobrain and experts for MH ratings was excellent with a single-measure intraclass correlation coefficient (ICC) of 0.89 (95% CI: 0.83-0.93). AI-derived MRI markers, particularly GM volume and MH burden, provide valuable predictors of cognitive response and ARIA risk in patients treated with lecanemab. Integrating quantitative neuroimaging into clinical workflows may enhance personalized treatment decisions and improve real-world implementation of Amyloid-targeting therapies.