Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RHSIA: Real-time Hemodynamics Surrogation for Non-idealized Intracranial Aneurysms

Yiying Sheng, Wenhao Ding, Dylan Roi, Leonard Leong Litt Yeo, Hwa Liang Leo, Choon Hwai Yap

TL;DR

This work targets the translation of CFD-derived intracranial aneurysm hemodynamics into clinical practice by introducing a real-time surrogate model. It presents a Graph Transformer framework that leverages Graph Harmonic Deformation encodings to map IA morphology to transient wall shear stress across the cardiac cycle, supervised by a large CFD dataset and augmented with cheap steady-flow data. The results show that steady-data augmentation substantially improves accuracy when pulsatile data are scarce, achieving high fidelity in WSS predictions and reliable computation of derived metrics such as TAWSS and RRT, with SSIM improvements up to 0.981 and rL2* around 2.8%. This approach promises real-time, patient-specific hemodynamics from geometry alone, with potential applicability to other cardiovascular scenarios and integration into clinical workflows for improved UIA risk assessment.

Abstract

Extensive studies suggested that fluid mechanical markers of intracranial aneurysms (IAs) derived from Computational Fluid Dynamics (CFD) can indicate disease progression risks, but to date this has not been translated clinically. This is because CFD requires specialized expertise and is time-consuming and low throughput, making it difficult to support clinical trials. A deep learning model that maps IA morphology to biomechanical markers can address this, enabling physicians to obtain these markers in real time without performing CFD. Here, we show that a Graph Transformer model that incorporates temporal information, which is supervised by large CFD data, can accurately predict Wall Shear Stress (WSS) across the cardiac cycle from IA surface meshes. The model effectively captures the temporal variations of the WSS pattern, achieving a Structural Similarity Index (SSIM) of up to 0.981 and a maximum-based relative L2 error of 2.8%. Ablation studies and SOTA comparison confirmed its optimality. Further, as pulsatile CFD data is computationally expensive to generate and sample sizes are limited, we engaged a strategy of injecting a large amount of steady-state CFD data, which are extremely low-cost to generate, as augmentation. This approach enhances network performance substantially when pulsatile CFD data sample size is small. Our study provides a proof of concept that temporal sequences cardiovascular fluid mechanical parameters can be computed in real time using a deep learning model from the geometric mesh, and this is achievable even with small pulsatile CFD sample size. Our approach is likely applicable to other cardiovascular scenarios.

RHSIA: Real-time Hemodynamics Surrogation for Non-idealized Intracranial Aneurysms

TL;DR

This work targets the translation of CFD-derived intracranial aneurysm hemodynamics into clinical practice by introducing a real-time surrogate model. It presents a Graph Transformer framework that leverages Graph Harmonic Deformation encodings to map IA morphology to transient wall shear stress across the cardiac cycle, supervised by a large CFD dataset and augmented with cheap steady-flow data. The results show that steady-data augmentation substantially improves accuracy when pulsatile data are scarce, achieving high fidelity in WSS predictions and reliable computation of derived metrics such as TAWSS and RRT, with SSIM improvements up to 0.981 and rL2* around 2.8%. This approach promises real-time, patient-specific hemodynamics from geometry alone, with potential applicability to other cardiovascular scenarios and integration into clinical workflows for improved UIA risk assessment.

Abstract

Extensive studies suggested that fluid mechanical markers of intracranial aneurysms (IAs) derived from Computational Fluid Dynamics (CFD) can indicate disease progression risks, but to date this has not been translated clinically. This is because CFD requires specialized expertise and is time-consuming and low throughput, making it difficult to support clinical trials. A deep learning model that maps IA morphology to biomechanical markers can address this, enabling physicians to obtain these markers in real time without performing CFD. Here, we show that a Graph Transformer model that incorporates temporal information, which is supervised by large CFD data, can accurately predict Wall Shear Stress (WSS) across the cardiac cycle from IA surface meshes. The model effectively captures the temporal variations of the WSS pattern, achieving a Structural Similarity Index (SSIM) of up to 0.981 and a maximum-based relative L2 error of 2.8%. Ablation studies and SOTA comparison confirmed its optimality. Further, as pulsatile CFD data is computationally expensive to generate and sample sizes are limited, we engaged a strategy of injecting a large amount of steady-state CFD data, which are extremely low-cost to generate, as augmentation. This approach enhances network performance substantially when pulsatile CFD data sample size is small. Our study provides a proof of concept that temporal sequences cardiovascular fluid mechanical parameters can be computed in real time using a deep learning model from the geometric mesh, and this is achievable even with small pulsatile CFD sample size. Our approach is likely applicable to other cardiovascular scenarios.
Paper Structure (24 sections, 11 equations, 8 figures, 6 tables)

This paper contains 24 sections, 11 equations, 8 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep Learning-based WSS Surrogation
  4. Geometric Graph Learning on Meshes
  5. Method
  6. Overview
  7. Shape Preprocessing
  8. Geometry Encoding
  9. Time and Waveform Encoding
  10. Augmentation with Cheap Steady Data
  11. Comparative Baselines
  12. Implementation Details
  13. Metrics
  14. Results
  15. Performance Comparison
  16. ...and 9 more sections

Figures (8)

  • Figure 1: Graph Harmonic Deformation (GHD) shape modes. Higher modes correspond to high-frequency behaviors on the graph.
  • Figure 2: Top: GHD-deformed mesh and the down-sampled versions. Bottom: KNN-constructed graphs where false connections are observed in highlighted regions.
  • Figure 3: Network schematic of our Graph Transformer.
  • Figure 4: Qualitative comparison of CFD and DL-predicted transient WSS field.
  • Figure 5: Qualitative comparison of CFD and DL-derived TAWSS, OSI, and RRT fields.
  • ...and 3 more figures