Dynamic Scatterer Decoupled Modeling and Real-Time Signal Generation Optimization for Complex Vasculature-Oriented Doppler Ultrasound Simulation

다국어 초록 (Multilingual Abstract)

In the non-invasive diagnosis of cardiovascular diseases, Doppler ultrasound technology provides critical quantitative support for cardiac valve assessment and vascular stenosis localization by detecting hemodynamic parameters such as flow velocity distribution. However, traditional simulation methods (e.g., time-domain correlation algorithms and finite element flow field interpolation) suffer from high angle sensitivity and computational complexity. Particularly in complex vascular scenarios, the efficient storage of dynamic scatterers and real-time signal generation via ray tracing have become core bottlenecks restricting simulation performance.
To address this, this study proposes a dynamic Doppler ultrasound simulation algorithm optimized for scatterers, with three core designs: first, constructing a vascular-scatterer decoupling framework, where blood flow is characterized by parameterized vascular meshes and random scatterers in bounding boxes; the parabolic model is used to define velocity, and the Gaussian distribution to describe scattering intensity, thus avoiding interference from tissue motion; second, designing a memory pool-hash mapping storage architecture, which achieves adaptive memory management through dynamic pre-allocation, hierarchical indexing, and load factors, reducing the overhead of scatterer construction and access; third, locating effective scatterer regions via contribution window sampling, and accelerating traversal and signal accumulation with CUDA (Compute Unified Device Architecture) parallelization to improve real-time performance.
Experimental results show that when processing 100,000-level scatterers, the construction time of the memory pool scheme is only 19% of that of traditional dynamic allocation, with memory usage increasing linearly; when processing 1,000,000-level scatterers, the time consumption after CUDA acceleration is 54 times less than that of CPU; the velocity estimation error is only 5.6%, meeting clinical accuracy requirements.
This study breaks through the technical bottlenecks in complex vascular scenarios. The proposed algorithm provides an extensible framework for high-precision Doppler ultrasound simulation and holds significant value for promoting the application of computer-aided diagnosis systems in clinical training and disease diagnosis.