Application of feed forward and recurrent neural networks in simulation of left ventricular mechanics

Abstract

An understanding of left ventricle (LV) mechanics is fundamental for designing better preventive, diagnostic, and treatment strategies for improved heart function. Because of the costs of clinical and experimental studies to treat and understand heart function, respectively, in-silico models play an important role. Finite element (FE) models, which have been used to create in-silico LV models for different cardiac health and disease conditions, as well as cardiac device design, are time-consuming and require powerful computational resources, which limits their use when real-time results are needed. As an alternative, we sought to use deep learning (DL) for LV in-silico modeling. We used 80 four-chamber heart FE models for feed forward, as well as recurrent neural network (RNN) with long short-term memory (LSTM) models for LV pressure and volume. We used 120 LV-only FE models for training LV stress predictions. The active material properties of the myocardium and time were features for the LV pressure and volume training, and passive material properties and element centroid coordinates were features of the LV stress prediction models. For six test FE models, the DL error for LV volume was 1.599 ± 1.227 ml, and the error for pressure was 1.257 ± 0.488 mmHg; for 20 LV FE test examples, the mean absolute errors were, respectively, 0.179 ± 0.050 for myofiber, 0.049 ± 0.017 for cross-fiber, and 0.039 ± 0.011 kPa for shear stress. After training, the DL runtime was in the order of seconds whereas equivalent FE runtime was in the order of several hours (pressure and volume) or 20 min (stress). We conclude that using DL, LV in-silico simulations can be provided for applications requiring real-time results.

Personalization of biomechanical simulations of the left ventricle by in-vivo cardiac DTI data: Impact of fiber interpolation methods

“…The computational cardiac biomechanics model was adapted from the Living Heart Human Project (version 2.1, Simulia, Dassault Systèmes) ( Baillargeon et al, 2014 ), previously used for biomechanical simulation studies by ( Genet et al, 2014 ; Sack et al, 2016a , b ; Genet et al, 2016 ; Sack et al, 2018b , a ; Peirlinck et al, 2018 ; Dabiri et al, 2018 ; Sahli Costabal et al, 2019 ; Dabiri et al, 2019b , a ; Peirlinck et al, 2021 ; Guan et al, 2020 ; Sack et al, 2020 ; Dabiri et al, 2020 ; Wisneski et al, 2020 ; Heidari et al, 2022 ; St. Pierre et al, 2022 ). The dynamic problem was solved in Abaqus 2020 (Simulia, Dassault Systèmes) using meshes of the LV with 160,000 linear tetrahedral elements with an average edge length of 1.8 ± 0.5 mm.…”

Mitral Valve Atlas for Artificial Intelligence Predictions of MitraClip Intervention Outcomes

“…On the other hand, ML models do not suffer from FE modeling limitations mentioned above. Once a ML model is developed, it can provide the results in much shorter time than FE modeling (25,28), and it can provide the results for the MV parameters that are within the data distribution (results will be provided except for outliers). In this paper, we described data generation, an important step in using ML analysis for MC outcomes prediction.…”

“…The LV deformations should be considered in future models where the contraction of the LV will cause the ejection of the blood. This future work can integrate our LV models (25,28) with our MV and MC modeling approach. The MV will be a part of the LV in such a way that the plate used to seal the gap between MV and LV will not be required.…”

A Deep Learning Model for the Identification of Active Contraction Properties of the Myocardium Using Limited Clinical Metrics

“…Informational models can interpret hidden patterns between input and output via repeated training on high-quality experimental data (Hoerig et al 2017). The fast and accurate predictions made by the deep learning model have been used in diverse elds of studies of cardiovascular mechanics (Dabiri et al 2020;Galati et al 2022; Kadem et al 2022;Liang et al 2018;Madani et al 2019). Moreover, emerging advancements in medical imaging techniques and computational modeling methods enable the generation of patientspeci c computational ventricular models (Litjens et al 2019;Miller et al 2021;Romaszko et al 2021; Tang et al 2010).…”