Investigation, if omitting subsets of channels in 32-electrode EIT data can improve performance and lower computational complexity of a CNN estimating aortic blood pressure from EIT
Abstract:
Electrical impedance tomography is a powerful monitoring tool that can be used to estimate
central aortic pressure with a convolutional neural network. This paper explores the selection
of subsets from the \num{1024} possible channels for reduced computational complexity and improved
network performance. The results show an improved performance for appropriate selection strategies.
This repository includes:
- create_cha_selection.py: To derive the indexing for different channel selection strategies
- get_xcdc_idx.py: To derive the XCDC-based indexing.
- plot_results.py: Final plotting functions.
- train_model1.py : To train the designed CNN for a specified channel selection strategy and save the network.
- eval_model.py: To reload the trained networks and visualize results.
Clone the repository:
git clone https://github.com/EITLabworks/EIT-Channel-Selection-for-Estimating-Aortic-Blood-Pressure.git
This repository is created by Patricia Fuchs, Institute of Communications Engineering, University of Rostock, Germany.
The research is explained and summarized in the paper "EIT Channel Selection for Estimating Aortic Blood Pressure" for
the "26th International Conference on Biomedical Applications of Electrical Impedance Tomography" (EIT) 2026.
For questions, please contact: pat.fuchs@uni-rostock.de