ResFouriONet: A Residual Fourier Operator Network with Synthetic Data Generation for Real-Time Laser-Induced Bioheat Transfer Modeling

ResFouriONet:一种用于实时激光诱导生物传热建模的具有合成数据生成的残差傅里叶算子网络

阅读:1

Abstract

Accurate modeling of laser-tissue interactions is essential for applications such as multiphoton brain imaging and laser-based therapeutics, where precise thermal control is critical. However, traditional numerical solvers for bioheat transfer, especially when paired with Monte Carlo (MC) simulations for light transport, are computationally prohibitive for real-time use. While operator learning offers a scalable alternative, existing architectures often struggle with extrapolatory tasks and rely on large, labeled datasets that are costly to generate. In this study, we introduce ResFouriONet, a residual deep operator network tailored for modeling 2D transient bioheat transfer with parametric laser sources. Our architecture features a dual-pathway branch network comprising a Fourier sub-branch for capturing global thermal patterns and a cascaded residual sub-branch for localized heating, fused via a self-attention mechanism. To reduce the reliance on expensive MC data, we introduce a synthetic laser source function generation scheme that combines parameterized Gaussian profiles, exponential decay, and random field perturbations to emulate the complexity of MC-derived fluence maps. ResFouriONet achieves a 5× reduction in prediction error compared to vanilla deep operator networks, despite having fewer parameters. It attains an average relative error of 0.93% on unseen synthetic data and 1.45% on 630 zero-shot MC-generated source functions. Together, the proposed architecture and data generation strategy enable accurate, real-time prediction of laser-induced heating, paving the way for simulation-free planning and control in biomedical imaging and laser-based therapies.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。