Graph convolutional networks enable fast hemorrhagic stroke monitoring with electrical impedance tomography

TL;DR

This work addresses fast online monitoring of intracerebral hemorrhage with electrical impedance tomography by combining a fast linear difference imaging (LD) reconstruction with a graph U-net post-processing step. The graph-based post-processing is trained on 2D cross-sections and applied to 3D Volumetric data, achieving image quality comparable to, or better than, the time-consuming nonlinear MO reconstructions while reducing data-simulation and computation costs by up to ~50x. The approach demonstrates strong generalization to out-of-distribution 3D head models and experimental data, offering substantial speedups (minutes vs hours) suitable for bedside monitoring. Overall, the LD + graph U-net pipeline provides a practical, scalable route toward real-time online monitoring of hemorrhagic stroke in clinical environments.

Abstract

Objective: To develop a fast image reconstruction method for stroke monitoring with electrical impedance tomography with image quality comparable to computationally expensive nonlinear model-based methods. Methods: A post-processing approach with graph convolutional networks is employed. Utilizing the flexibility of the graph setting, a graph U-net is trained on linear difference reconstructions from 2D simulated stroke data and applied to fully 3D images from realistic simulated and experimental data. An additional network, trained on 3D vs. 2D images, is also considered for comparison. Results: Post-processing the linear difference reconstructions through the graph U-net significantly improved the image quality, resulting in images comparable to, or better than, the time-intensive nonlinear reconstruction method (a few minutes vs. several hours). Conclusion: Pairing a fast reconstruction method, such as linear difference imaging, with post-processing through a graph U-net provided significant improvements, at a negligible computational cost. Training in the graph framework vs classic pixel-based setting (CNN) allowed the ability to train on 2D cross-sectional images and process 3D volumes providing a nearly 50x savings in data simulation costs with no noticeable loss in quality. Significance: The proposed approach of post-processing a linear difference reconstruction with the graph U-net could be a feasible approach for on-line monitoring of hemorrhagic stroke.

Figures (8)

• Figure 1: Demonstration of the proposed method. Fast, blurry linear difference (LD) reconstructions are post-processed with a graph U-net, at negligible cost, resulting in images of comparable quality to computationally expensive nonlinear methods.
• Figure 2: Graph convolutional networks can be trained on 2D or 3D input data. The graph U-net is trained on linear difference (LD) reconstructions, and then applied to linear difference images from simulated and experimental voltages. The resulting processed reconstruction is shown on the right.
• Figure 3: Results for processing simulated LD reconstructions, consistent with training/validation data, through the LD-gUnet2D network.
• Figure 4: Estimated probability density functions and their means for the error metrics for the original linear difference imaging reconstructions (LD), the reference stroke monitoring reconstructions (MO), and for the 2D and 3D network processed linear difference imaging samples (gUnet2d, top row of images; gUnet3d, bottom row of images) using data consistent with training and validation.
• Figure 5: Computational head model test case: Results for processing the linear difference imaging reconstructions (LD) through the 2D and 3D graph-Unet networks (LD-gUnet2D and LD-gUnet3D), and the reference stroke monitoring reconstructions (MO).