Predicting Plaque Vulnerability Using Machine Learning-Enabled Nanoagents Sensitized Molecular High-Resolution...

Area of Science:

• Biomedical imaging
• Nanotechnology
• Machine learning

Background:

• Molecular imaging with paramagnetic nanoagents shows promise for detecting atherosclerotic plaque destabilization.
• Current magnetic resonance (MR) imaging lacks quantitative descriptors for precise plaque risk stratification.

Purpose of the Study:

• To develop a novel nano-agent assisted machine learning (nano-AML) technology for direct plaque vulnerability assessment.
• To improve the noninvasive prediction of plaque vulnerability using molecular high-resolution vessel wall MR imaging (HR-VWI).

Main Methods:

• Developed synergistic nanoagents (tFM-Nanoagents) for enhanced MR imaging.
• Applied a machine learning (ML) approach to decode tFM-Nanoagents sensitized HR-VWI data.
• Generated and validated an imaging-derived risk score (nano-AML score) in a preclinical model.

Main Results:

• The nano-AML score effectively classified plaques as "vulnerable" or "stable" with high accuracy (AUC training: 0.871, validation: 0.870).
• The nano-AML score significantly outperformed the commercial contrast agent Gadovist in predicting plaque vulnerability (AUC training: 0.560, validation: 0.538).

Conclusions:

• The developed nano-AML technology provides a robust and reliable method for predicting vulnerable plaques.
• This approach has the potential to significantly advance the noninvasive assessment of atherosclerotic plaque risk.