This paper aims to demonstrate how electrical impedance tomography (EIT), with its functional imaging benefit, makes early diagnosis of lung cancer possible. To identify the location of tumor, various methods based on image reconstruction have been developed. However, due to the complexity of the detected object and the ill-posedness of the inverse problem, accurate image reconstruction is a great challenge. As a result, image-based identification accuracy raises a concern and a new method is required.
This paper develops an intelligent identification method that combines variational mode decomposition (VMD), convolutional neural network and bidirectional long short-term memory network. The measured signal is decomposed by VMD, and features of the optimal decomposed component are extracted. The extracted features are then input into a network model for identifying different cases. The identification results of the proposed method are compared with other six methods under the noiseless condition. Moreover, various factors including the effects of noise interference, conductivity variation and contact impedance change on the proposed method are analyzed.
The results demonstrate that the proposed approach shows the best identification accuracy. In contrast to the comparison methods, it is less sensitive to noise interference, conductivity variation and contact impedance change in identifying different cases.
The identification of tumor is implemented by direct analysis of boundary measurement. There is no need to solve the ill-posed inverse problem that is required in the traditional image-based identification approaches. A new choice for locating tumor with the lung EIT is provided.
