PEUNet:一种用于脑出血病灶检测的多频电阻抗成像方法

PEUNet:a multi-frequency electrical impedance tomography imaging method for the detection of cerebral hemorrhage lesions

目的提出一种新的多频电阻抗成像方法,实现颅脑模型中的出血目标成像。方法提出一种结合预处理模块和改进的UNet模型的PEUNet成像方法,首先通过预处理模块进行病灶图像初始化,随后采用注意力模块改进UNet模型,使其聚焦于图像重要特征,并用于初始图像的后处理,实现颅内出血目标的成像。结果在构建的多层颅脑仿真模型中,PEUNet方法准确地重构出了多种目标分布方式下的出血目标。结论在仿真模型上实现了多层颅脑模型中出血目标的成像,重构目标与真实目标的位置和面积差异较小,为后续多频EIT颅脑成像模型实验研究奠定了良好基础。

Objective To propose a new multi-frequency electrical impedance tomography imaging method to realize the imaging of hemorrhage targets in brain model.Methods A PEUNet imaging method combining preprocessing module and improved UNet model was proposed.Firstly,the preprocessing module was used to initialize the lesion image,and then the attention module was used to improve the UNet model so that it focused on the critical features of the image and was used for the post-processing of the initial image to realize the imaging of intracranial hemorrhage targets.Results In the constructed multi-layer brain simulation model,PEUNet method accurately reconstructed the hemorrhage targets in various target distribution modes.Conclusion On the simulation model,the imaging of hemorrhage targets in the multi-layer brain model is realized,and the location and area differences between the reconstructed target and the actual target are negligible,which lays a solid foundation for the follow-up experimental research of multi-frequency EIT brain imaging model.