Subject-specific simulation for non-invasive assessment of aortic coarctation: Towards a translational approach

Deep-learning-based real-time individualization for reduce-order haemodynamic model

The reduced-order lumped parameter model (LPM) has great computational efficiency in real-time numerical simulations of haemodynamics but is limited by the accuracy of patient-specific computation. This study proposed a method to achieve the individual LPM modeling with high accuracy to improve the practical clinical applicability of LPM. Clinical data was collected from two medical centres comprising haemodynamic indicators from 323 individuals, including brachial artery pressure waveforms, cardiac output data, and internal carotid artery flow waveforms. The data were expanded to 5000 synthesised cases that all fell within the physiological range of each indicator. LPM of the human blood circulation system was established. A double-path neural network (DPNN) was designed to input the waveforms of each haemodynamic indicator and their key features and then output the individual parameters of the LPM, which was labelled using a conventional optimization algorithm. Clinically collected data from the other 100 cases were used as the test set to verify the accuracy of the individual LPM parameters predicted by DPNN. The results show that DPNN provided good convergence in the training process. In the test set, compared with clinical measurements, the mean differences between each haemodynamic indicator and the estimate calculated by the individual LPM based on the DPNN were about 10 %. Furthermore, DPNN prediction only takes 4 s for 100 cases. The DPNN proposed in this study permits real-time and accurate individualization of LPM's. When facing medical issues involving haemodynamics, it lays the foundation for patient-specific numerical simulation, which may be beneficial for potential clinical application.

Deep learning based assessment of hemodynamics in the coarctation of the aorta: comparison of bidirectional recurrent and convolutional neural networks

The utilization of numerical methods, such as computational fluid dynamics (CFD), has been widely established for modeling patient-specific hemodynamics based on medical imaging data. Hemodynamics assessment plays a crucial role in treatment decisions for the coarctation of the aorta (CoA), a congenital heart disease, with the pressure drop (PD) being a crucial biomarker for CoA treatment decisions. However, implementing CFD methods in the clinical environment remains challenging due to their computational cost and the requirement for expert knowledge. This study proposes a deep learning approach to mitigate the computational need and produce fast results. Building upon a previous proof-of-concept study, we compared the effects of two different artificial neural network (ANN) architectures trained on data with different dimensionalities, both capable of predicting hemodynamic parameters in CoA patients: a one-dimensional bidirectional recurrent neural network (1D BRNN) and a three-dimensional convolutional neural network (3D CNN). The performance was evaluated by median point-wise root mean square error (RMSE) for pressures along the centerline in 18 test cases, which were not included in a training cohort. We found that the 3D CNN (median RMSE of 3.23 mmHg) outperforms the 1D BRNN (median RMSE of 4.25 mmHg). In contrast, the 1D BRNN is more precise in PD prediction, with a lower standard deviation of the error (±7.03 mmHg) compared to the 3D CNN (±8.91 mmHg). The differences between both ANNs are not statistically significant, suggesting that compressing the 3D aorta hemodynamics into a 1D centerline representation does not result in the loss of valuable information when training ANN models. Additionally, we evaluated the utility of the synthetic geometries of the aortas with CoA generated by using a statistical shape model (SSM), as well as the impact of aortic arch geometry (gothic arch shape) on the model's training. The results show that incorporating a synthetic cohort obtained through the SSM of the clinical cohort does not significantly increase the model's accuracy, indicating that the synthetic cohort generation might be oversimplified. Furthermore, our study reveals that selecting training cases based on aortic arch shape (gothic versus non-gothic) does not improve ANN performance for test cases sharing the same shape.

The effect of coarctation degrees on wall shear stress indices

Coarctation of the aorta (CoA) is a congenital tightening of the proximal descending aorta. Flow quantification can be immensely valuable for an early and accurate diagnosis. However, there is a lack of appropriate diagnostic approaches for a variety of cardiovascular diseases, such as CoA. An accurate understanding of the disease depends on measurements of the global haemodynamics (criteria for heart function) and also the local haemodynamics (detailed data on the dynamics of blood flow). Playing a significant role in clinical processes, wall shear stress (WSS) cannot be measured clinically; thus, computation tools are needed to give an insight into this crucial haemodynamic parameter. In the present study, in order to enable the progress of non-invasive approaches that quantify global and local haemodynamics for different CoA severities, innovative computational blueprint simulations that include fluid-solid interaction models are developed. Since there is no clear approach for managing the CoA regarding its severity, this study proposes the use of WSS indices and pressure gradient to better establish a framework for treatment procedures in CoA patients with different severities. This provides a platform for improving CoA therapy on a patient-specific level, in which physicians can perform treatment methods based on WSS indices on top of using a mere experience. Results show how severe CoA affects the aorta in comparison to the milder cases, which can give the medical community valuable information before and after any intervention.