# Importance of localized dilatation and distensibility in identifying determinants of thoracic aortic aneurysm with neural operators

## TL;DR

This study shows that combining shape and mechanical data improves the identification of causes behind thoracic aortic aneurysms, potentially leading to better personalized treatments.

## Contribution

The study introduces a novel approach using neural operators to integrate dilatation and distensibility data for predicting TAA causes.

## Key findings

- Prediction errors are significantly lower when using both dilatation and distensibility data compared to dilatation alone.
- UNet is identified as the best-performing neural network architecture for this task.
- Full-field measurements of dilatation and distensibility are crucial for identifying the underlying pathologic mechanisms of TAAs.

## Abstract

Thoracic aortic aneurysms (TAAs) stem from diverse mechanical and mechanobiological disruptions to the aortic wall that can also increase the risk of dissection or rupture. There is increasing evidence that dysfunctions along the aortic mechanotransduction axis, including reduced integrity of elastic fibers and loss of cell-matrix connections, are particularly capable of causing thoracic aortopathy. Because different insults can produce distinct mechanical vulnerabilities, there is a pressing need to identify interacting factors that drive progression. In this work, we employ a finite element framework to generate synthetic TAAs arising from hundreds of heterogeneous insults that span a range of compromised elastic fiber integrity and cellular mechanosensing. From these simulations, we construct localized dilatation and distensibility maps throughout the aortic domain to serve as training data for neural network models to predict the initiating combined insult. Several candidate architectures (Deep Operator Networks, UNets, and Laplace Neural Operators) and input data formats are compared to establish a standard for handling future subject-specific information. We further quantify the predictive capability when networks are trained on geometric (dilatation) information alone, which mimics current clinical guidelines, versus training on both geometric and mechanical (distensibility) information. We show that prediction errors based on dilatation data are significantly higher than those based on dilatation and distensibility across all networks considered, highlighting the benefit of obtaining local distensibility measures in TAA assessment. Additionally, we identify UNet as the best-performing architecture across all training data formats. These findings demonstrate the importance of obtaining full-field measurements of both dilatation and distensibility in the aneurysmal aorta to identify the mechanobiological insults that drive disease progression, which will advance personalized treatment strategies that target the underlying pathologic mechanisms.

We investigated causes of thoracic aortic aneurysms (TAAs), which are local enlargements of the aorta that can lead to life-threatening rupture. TAAs result from a combination of structural and cellular disruptions in the aortic wall, but need not pose the same risk even if appearing similar in size. We created computer models of aneurysms caused by different combinations of known risk factors, including tissue damage and cellular dysfunction. Using these simulations, we trained multiple neural network models with information on TAA shape (dilatation) and mechanics (distensibility) to see if they could recover the original causes. We found that predictions based on both dilatation and distensibility were significantly better than using dilatation alone, which represents the current clinical standard. Among the models tested, we also identified a best performing architecture, UNet, for these applications. These results suggest that measuring the full shape and mechanics of the aneurysmal aorta could lead to improved personalized diagnoses and treatment for TAAs.