Predicting and understanding arterial elasticity from key microstructural features by bidirectional deep learning

• DOI: 10.1016/j.actbio.2022.05.039
• Publication Date: 2022-05-25T14:45:40Z
• Authors: Kevin Linka; Cristina Cavinato; Jay D. Humphrey; Christian J. Cyron
• Keywords: Mice; Deep Learning; 0206 medical engineering; Animals; Arteries; Collagen; Stress, Mechanical; 02 engineering and technology; Elasticity; Biomechanical Phenomena; Extracellular Matrix

Microstructural features and mechanical properties are closely related in all soft biological tissues. Both yet exhibit considerable inter-individual differences affected by factors such as aging disease its progression. Histological analysis, modern situ imaging, biomechanical testing have deepened our understanding of these complex interrelations, two key questions remain: (1) Given the specific microstructure, can one predict macroscopic without testing? (2) Can quantify individual contributions different microstructural to an automated, systematic largely unbiased way? Here we propose a bidirectional deep learning architecture address questions. Our uses data from standard histological analyses, two-photon microscopy biaxial testing. Its capabilities demonstrated predicting with high accuracy (R2=0.92) evolving murine aorta during maturation aging. Moreover, reveals that extracellular matrix composition organization most prominent governing tissues studied herein. STATEMENT OF SIGNIFICANCE: We present physics-informed machine arterial tissue microstructure (characterized imaging data). For first time, this enables also fully automatic quantification relevance (such collagen volume fraction fiber straightness) for properties. This approach opens up unprecedented ways predictive modeling it provides quantitative insights into relation between promise play important role future engineering.