AI for ARIA Detection: How icobrain aria Supports Alzheimer’s Treatment Monitoring
Updated: 14 hours ago
As disease-modifying treatments for Alzheimer’s disease become more widely used, monitoring for amyloid-related imaging abnormalities (ARIA) has become an increasingly important part of clinical care. Detecting and assessing ARIA on MRI requires careful review, particularly as findings can influence treatment continuation, dose modification and patient care decisions.
A new expert opinion published in the American Journal of Roentgenology (AJR) explores how artificial intelligence can support ARIA detection and its responsible integration into clinical practice. The publication, led by Jeffrey R. Petrella, MD, of Duke University, brings together nine experts in neuroradiology and Alzheimer’s disease, including Frederik Barkhof (University College London and Amsterdam UMC), Tammie Benzinger (Washington University in St. Louis), Stephen Salloway (Brown University) and Greg Zaharchuk (Stanford University).
Among the four commercially available AI tools evaluated in the publication, icobrain aria* stands out for its specific intended use in ARIA detection and assessment, including severity rating.
A distinct role for icobrain aria among FDA-cleared tools
The expert panel evaluated four commercially available AI tools with ARIA-related functionality. All four have received FDA 510(k) clearance, but their intended uses and clinical capabilities differ.
icobrain aria* is the only one of the four AI tools evaluated by the panel that is specifically intended for ARIA detection and assessment. Its capabilities include identifying ARIA-E and ARIA-H, documenting their location and providing severity ratings. The other evaluated tools offer broader lesion-analysis or imaging-analysis functionality with ARIA-relevant outputs, but do not have the same ARIA-specific intended use or severity-rating functionality described in the publication.
This distinction can be important clinically. ARIA assessment involves more than identifying abnormalities. Severity classification is relevant to decisions about monitoring and the continuation or interruption of Alzheimer’s treatment. By combining detection, quantification and severity assessment in a dedicated ARIA solution, icobrain aria* addresses this specific clinical requirement.
Clinical evidence supporting AI-assisted ARIA assessment
The AJR publication reviews the available validation evidence for the four evaluated tools. For icobrain aria*, it highlights a peer-reviewed reader study by Sima et al., published in JAMA Network Open in 2024.¹
The study included 199 patients treated with aducanumab, with paired baseline and follow-up MRI examinations, and 16 radiologists who assessed the scans with and without AI assistance. Three expert neuroradiologists established the reference standard for ARIA presence and severity.
With AI assistance, radiologists achieved improved performance in detecting both ARIA-E and ARIA-H, alongside shorter median reading times. The study reported an area under the curve (AUC) of 0.873 for ARIA-E and 0.825 for ARIA-H. Sensitivity and specificity were 86.5% and 83.0% for ARIA-E, and 79.0% and 80.3% for ARIA-H, respectively.¹
These findings provide clinical evidence for the role of icobrain aria* in supporting radiologists' ARIA assessments. Importantly, they demonstrate the potential of AI assistance while keeping the radiologist responsible for the final interpretation.
What leading experts say about AI in ARIA monitoring
The AJR expert panel concludes that AI-assisted ARIA detection is likely to help enhance patient safety when used as clinical decision support within a radiologist-in-the-loop framework.
The panel recognises that missed ARIA findings and false-positive results do not have equivalent clinical consequences. Missing clinically significant ARIA may compromise patient safety, whereas false-positive findings can lead to additional imaging, further interpretation and potentially unnecessary treatment modifications.
The experts therefore support conditional implementation of AI-assisted ARIA detection, provided that appropriate safeguards are in place. These include:
Radiologist oversight and responsibility for final interpretation.
Local validation before clinical implementation.
Standardised MRI acquisition and appropriate reader training.
Ongoing quality assurance and prospective monitoring of clinical performance.
The panel also emphasises that better detection does not necessarily translate into improved clinical outcomes. Further prospective research is needed to establish the potential impact of AI-assisted ARIA monitoring on treatment decisions and patient safety.
From detection to clinically relevant severity assessment
One of the key considerations in ARIA monitoring is distinguishing between findings of different severity. ARIA-E, in particular, can influence decisions about continuing or interrupting anti-amyloid treatment. Consistent assessment is therefore important throughout a patient's treatment journey.
icobrain aria* combines ARIA detection with quantitative reporting and severity ratings for both ARIA-E and ARIA-H. Its longitudinal capabilities also support the assessment of findings across MRI examinations.
This dedicated functionality distinguishes icobrain aria* within the group of evaluated FDA-cleared tools. Rather than providing only ARIA-relevant lesion outputs, it is specifically intended to support ARIA detection and assessment, including severity classification.
Responsible AI integration keeps radiologists in control
The expert opine that introducing AI into ARIA monitoring requires more than regulatory. Validation, appropriate training, standardised imaging protocols and continuous quality monitoring are essential to responsible clinical implementation.
For radiology departments, integration into existing reporting practices is also an important consideration. AI should support the interpretation process without displacing radiologists' clinical judgement or adding unnecessary complexity.
icobrain aria* is designed specifically for ARIA assessment, providing structured, quantitative information to support radiologists in their reporting. Its FDA 510(k) clearance and peer-reviewed validation evidence provide a foundation for its role in clinical practice.
Advancing ARIA monitoring with dedicated AI support
The AJR expert opinion highlights both the potential of AI-assisted ARIA detection and the importance of responsible clinical implementation. It also identifies meaningful differences between commercially available tools in their intended uses, capabilities and supporting evidence.
For icometrix, the publication provides important independent context for the role of icobrain aria*. Its specific intended use for ARIA detection and assessment, including severity ratings, distinguishes it from the other evaluated tools, while published clinical evidence supports its use as an aid to radiologists.
As Alzheimer’s treatment monitoring evolves, icobrain aria* is positioned to support structured ARIA assessment while keeping clinical judgement where it belongs: with the radiologist.
Read the full publication: Artificial Intelligence–Based Detection of ARIA on MRI During Alzheimer Disease Therapy: Expert Opinion on Responsible Clinical Integration
*icobrain aria is 510(k) cleared. CE-marked as part of icobrain mr. Not available in all markets.
Sima DM, Phan TV, Van Eyndhoven S, et al. Artificial Intelligence Assistive Software Tool for Automated Detection and Quantification of Amyloid-Related Imaging Abnormalities. JAMA Network Open. 2024;7(2):e2355800. doi:10.1001/jamanetworkopen.2023.55800.
