Yi Huangjian, Yang Ruigang, Wang Yishuo, Wang Yihan, Guo Hongbo, Cao Xu, Zhu Shouping, He Xiaowei
School of Information Sciences and Technology, Northwest University, Xi'an, Shaanxi 710069, China.
The Xi'an Key Laboratory of Radiomics and Intelligent Perception, No. 1 Xuefu Avenue, 710127 Xi'an, Shaanxi, China.
Biomed Opt Express. 2024 Feb 27;15(3):1910-1925. doi: 10.1364/BOE.509775. eCollection 2024 Mar 1.
Diffuse optical tomography (DOT) employs near-infrared light to reveal the optical parameters of biological tissues. Due to the strong scattering of photons in tissues and the limited surface measurements, DOT reconstruction is severely ill-posed. The Levenberg-Marquardt (LM) is a popular iteration method for DOT, however, it is computationally expensive and its reconstruction accuracy needs improvement. In this study, we propose a neural model based iteration algorithm which combines the graph neural network with Levenberg-Marquardt (GNNLM), which utilizes a graph data structure to represent the finite element mesh. In order to verify the performance of the graph neural network, two GNN variants, namely graph convolutional neural network (GCN) and graph attention neural network (GAT) were employed in the experiments. The results showed that GCNLM performs best in the simulation experiments within the training data distribution. However, GATLM exhibits superior performance in the simulation experiments outside the training data distribution and real experiments with breast-like phantoms. It demonstrated that the GATLM trained with simulation data can generalize well to situations outside the training data distribution without transfer training. This offers the possibility to provide more accurate absorption coefficient distributions in clinical practice.
扩散光学层析成像（DOT）利用近红外光来揭示生物组织的光学参数。由于光子在组织中的强烈散射以及有限的表面测量，DOT重建是严重不适定的。Levenberg-Marquardt（LM）方法是一种用于DOT的常用迭代方法，然而，它计算成本高且重建精度有待提高。在本研究中，我们提出了一种基于神经模型的迭代算法，该算法将图神经网络与Levenberg-Marquardt（GNNLM）相结合，它利用图数据结构来表示有限元网格。为了验证图神经网络的性能，实验中采用了两种GNN变体，即图卷积神经网络（GCN）和图注意力神经网络（GAT）。结果表明，在训练数据分布内的模拟实验中，GCNLM表现最佳。然而，在训练数据分布外的模拟实验以及使用类乳腺体模的实际实验中，GATLM表现出卓越的性能。这表明用模拟数据训练的GATLM无需迁移训练就能很好地推广到训练数据分布之外的情况。这为在临床实践中提供更准确的吸收系数分布提供了可能性。
