Application of Machine Learning for Volumetric Analysis of Atherosclerotic Burden

Cardiovascular disease (CVD) remains the leading cause of mortality worldwide. Preclinical studies to research and validate therapeutic interventions for CVD often depend on two- dimensional histological surveys. The use of light sheet fluorescence microscopy together with optical clearing methods amenable to immunofluorescence staining are recent advances, all of which deliver detailed three-dimensional rendering of vessels. This offers the ability to describe and quantify features critical in CVD models, specifically, atherosclerotic plaque burden in atherosclerotic animal models and neointimal hyperplasia in surgical models. The main challenge for this approach remains the lengthy, hands-on, analysis time. Labkit is a user- friendly Fiji plugin that applies a machine-learning algorithm to create 3D renderings from large microscopy data. Likewise, syGlass a virtual reality (VR) software, allows for 3D visualization and analysis of information-rich image datasets. The application of these tools is expected to decrease the hands-on analysis time required to generate accurate volumetric renderings of arterial disease and injury features in animal models of CVD. For atherosclerotic burden analysis, Ldlr-/- (C57/BL6) mice aged 6-8 weeks were fed a high-fat diet for 15 weeks to allow the development of atherosclerotic plaque along the aorta. For neointimal hyperplasia analysis, surgically intervened carotid arteries from rats and mice were collected 2 weeks post-surgery. iDISCO+ or AdipoClear and immunolabeling together with light-sheet fluorescence microscopy allowed for three-dimensional visualization of the vessels. Both Imaris software v9.9.1 and the built-in bridge to ImageJ/Labkit were used to quantify plaque burden and neointimal hyperplasia manually or automatically. syGlass was also utilized for the quantification of plaque burden and other disease-associated characteristics. Our findings indicate that both Labkit and syGlass offer effective and user-friendly platforms for the segmentation of atherosclerotic plaque and/or neointimal hyperplasia in animal models.