The Potential of Deep Learning to Advance Clinical Applications of Computational Biomechanics

Abstract

When combined with patient information provided by advanced imaging techniques, computational biomechanics can provide detailed patient-specific information about stresses and strains acting on tissues that can be useful in diagnosing and assessing treatments for diseases and injuries. This approach is most advanced in cardiovascular applications but can be applied to other tissues. The challenges for advancing computational biomechanics for real-time patient diagnostics and treatment include errors and missing information in the patient data, the large computational requirements for the numerical solutions to multiscale biomechanical equations, and the uncertainty over boundary conditions and constitutive relations. This review summarizes current efforts to use deep learning to address these challenges and integrate large data sets and computational methods to enable real-time clinical information. Examples are drawn from cardiovascular fluid mechanics, soft-tissue mechanics, and bone biomechanics. The application of deep-learning convolutional neural networks can reduce the time taken to complete image segmentation, and meshing and solution of finite element models, as well as improving the accuracy of inlet and outlet conditions. Such advances are likely to facilitate the adoption of these models to aid in the assessment of the severity of cardiovascular disease and the development of new surgical treatments.

Keywords: biomechanics; deep learning; finite element models; image segmentation; neural networks.

Figure 1

Schematic of the steps in the simulation of patient-specific biomechanics.

Figure 2

Schematic of neural networks for prediction of fluid velocity and pressure from training sets consisting of experimental measurements and computer simulations. The green circles represent the different layers of the neural network. The red circles represent the derived parameters (v_x, vy, v_z, and p) from the neural network.

Figure 3

Schematic of physics-informed neural networks (PINN) applied to a solution of the Navier–Stokes equation and boundary conditions. I represents the identity matrix, n is the normal unit surface facing into the fluid. The green circles represent the different layers of the neural network. The red circles represent the different operations performed on the derived parameters (v_x, vy, v_z, and p) from the neural network.

Figure 4

(A) Streamlines at low resolution through the right and left internal and middle carotid artery are similar between original and enhanced flow fields. A higher-magnification view of the internal carotid artery bend in (B) shows that streamlines from the original flow field tend to point and terminate outside the vessel wall, while streamlines in the enhanced flow field retain normal flow paths. From [62] and published based on a CC-BY Creative Commons license from Scientific Reports.

Figure 5

Computational process using neural networks to map the fiber structure to mesh for isogeometric analysis (IGA) model of bioprosthetic heart valves. E—Green–Lagrange strain tensor; σ_θ—standard deviation of preferred fiber direction; S—second Piola–Kirchhoff stress tensor (Equation (3)). Reprinted from [70], with permission.