Deep learning reveals 3D atherosclerotic plaque distribution and composition.

Area of Science:

• Cardiovascular Research
• Medical Imaging
• Computational Biology

Background:

• Atherosclerosis complications are a major global health burden.
• Classical histology is the standard for analyzing atherosclerotic plaque composition.
• Genetically modified mouse models are crucial for studying atherosclerosis progression.

Purpose of the Study:

• To demonstrate the utility of light-sheet fluorescence microscopy and deep learning for characterizing and quantifying atherosclerotic plaque in whole mouse aortas.
• To establish a non-destructive, scalable imaging method for atherosclerosis research.
• To investigate the impact of anatomical location on plaque composition and progression.

Main Methods:

• Utilized light-sheet fluorescence microscopy for 3D imaging of whole aorta specimens.
• Applied deep learning algorithms for automated identification and quantification of atherosclerotic plaque without ex vivo staining.
• Segmented aorta and its branches to analyze anatomical variations in plaque burden.
• Quantified immune cell infiltration using CD45 staining in ApoE-/- mice.

Main Results:

• Identified highest plaque accumulation in the aortic arch and brachiocephalic artery.
• Demonstrated that atherosclerotic plaque can be identified via autofluorescence, reducing the need for staining.
• Observed that in ApoE-/- mice, CD45 levels plateau, indicating plaque volume increases may not correlate with immune cell infiltration beyond a certain point.
• Developed and shared open-source code for method replication.

Conclusions:

• Light-sheet fluorescence microscopy combined with deep learning offers a powerful, non-destructive approach for atherosclerosis research.
• This method enables detailed analysis of plaque burden, composition, and anatomical distribution.
• The findings highlight a dissociation between plaque volume and immune cell infiltration in later stages of the disease in ApoE-/- mice.