Related Experiment Video
Updated: Sep 13, 2025
Visualization of Amyloid β Deposits in the Human Brain with Matrix-assisted Laser Desorption/Ionization Imaging Mass Spectrometry
Published on: March 7, 2019
Artificial Intelligence-Assisted Detection of Amyloid-Related Imaging Abnormalities: Promise and Pitfalls
Jeffrey R Petrella1,2, Andrew J Liu3,2, Laura A Wang4
1From the Department of Radiology (J.R.P.), Alzheimer's Imaging Research Lab, Duke University School of Medicine, Durham, North Carolina jeffrey.petrella@duke.edu.
Abstract:
The advent of anti-amyloid therapies (AATs) for Alzheimer disease (AD) has elevated the importance of MRI surveillance for amyloid-related imaging abnormalities (ARIA) such as microhemorrhages and siderosis (ARIA-H) and edema (ARIA-E). We report a literature review and early quality-assurance experience with an FDA-cleared decision support artificial intelligence (AI) tool intended for detection of ARIA in MRI clinical workflows. The AI system improved the sensitivity for detection of subtle ARIA-E and ARIA-H lesions but at the cost of a reduction in specificity. We propose a tiered workflow combining protocol harmonization and expert interpretation with an AI overlay review. AI-assisted ARIA detection is a paradigm shift that offers great promise to enhance patient safety as disease-modifying therapies for AD gain broader clinical use; however, some pitfalls need to be considered.
Insights
Anti-amyloid therapies for Alzheimer's disease require MRI monitoring for imaging abnormalities. An AI tool improved detection of these abnormalities but reduced specificity, necessitating a combined human-AI approach for patient safety.
Area of Science:
- Neurology
- Radiology
- Artificial Intelligence
Background:
- Anti-amyloid therapies (AATs) for Alzheimer's disease (AD) necessitate vigilant MRI surveillance for amyloid-related imaging abnormalities (ARIA).
- ARIA encompasses microhemorrhages and siderosis (ARIA-H) and edema (ARIA-E), critical indicators requiring accurate detection.
- The increasing use of AATs highlights the need for efficient and reliable ARIA detection methods in clinical practice.
Purpose of the Study:
- To review the literature and evaluate the early quality assurance experience of an FDA-cleared AI tool for ARIA detection in MRI workflows.
- To assess the performance of AI in identifying subtle ARIA-E and ARIA-H lesions.
- To propose an optimized workflow integrating AI assistance for ARIA detection.
Main Methods:
- Literature review on ARIA detection in AD patients undergoing AATs.
- Quality assurance assessment of an AI tool designed for detecting ARIA-E and ARIA-H on MRI scans.
- Analysis of AI performance metrics, including sensitivity and specificity.
Main Results:
- The AI system demonstrated improved sensitivity for detecting subtle ARIA-E and ARIA-H lesions.
- A trade-off was observed, with increased sensitivity accompanied by a reduction in specificity.
- Early experience suggests AI-assisted detection is promising but requires careful implementation.
Conclusions:
- AI-assisted ARIA detection represents a significant advancement for patient safety in the context of emerging AATs for AD.
- A proposed tiered workflow combines protocol harmonization, expert interpretation, and AI overlay review for optimal results.
- While promising, potential pitfalls of AI implementation in ARIA detection must be addressed for widespread clinical adoption.
