Neural operator learning of heterogeneous mechanobiological insults contributing to aortic aneurysms

• DOI: 10.1098/rsif.2022.0410
• Publication Date: 2022-08-31T07:05:44Z
• Authors: Somdatta Goswami; David S. Li; Bruno V. Rego; Marcos Latorre; Jay D. Humphrey; George Em Karniad...
• Keywords: FOS: Computer and information sciences; 0301 basic medicine; Computer Science - Machine Learning; Aortic Aneurysm, Thoracic; Biophysics; Deep learning; Quantitative Biology - Tissues and Organs; Growth and remodelling; Biomechanical Phenomena; Machine Learning (cs.LG); 03 medical and health sciences; Risk Factors; Thoracic aortic aneurysm; FOS: Biological sciences; Operator-based neural network; 03.- Garantizar una vida saludable y promover el bienestar para todos y todas en todas las edades; Humans; Tissues and Organs (q-bio.TO); Aorta

Thoracic aortic aneurysm (TAA) is a localized dilatation of the aorta that can lead to life-threatening dissection or rupture. In vivo assessments TAA progression are largely limited measurements size and growth rate. There promise, however, computational modelling evolving biomechanics could predict future geometry properties from initiating mechanobiological insults. We present an integrated framework train deep operator network (DeepONet)-based surrogate model identify contributing factors using synthetic finite-element-based datasets. For training, we employ constrained mixture remodelling generate maps local distensibility for multiple risk factors. evaluate performance insult distributions varying fusiform (analytically defined) complex (randomly generated). propose two frameworks, one trained on sparse information full-field greyscale images, gain insight into preferred neural operator-based approach. show this continuous learning approach patient-specific profile associated with any given map high accuracy, particularly when based images. Our findings demonstrate feasibility applying DeepONet support transfer inputs progression.