RESUMO
Objective.Electrical impedance tomography is a valuable tool for monitoring global and regional lung mechanics. To evaluate the recorded data, an accurate estimate of the lung area is crucial.Approach.We present two novel methods for estimating the lung area using functional tidal images or active contouring methods. A convolutional neural network was trained to determine, whether or not the heart region was visible within tidal images. In addition, the effects of lung area mirroring were investigated. The performance of the methods and the effects of mirroring were evaluated via a score based on the impedance magnitudes and their standard deviations in functional tidal images.Main results.Our analyses showed that the method based on functional tidal images provided the best estimate of the lung area. Mirroring of the lung area had an impact on the accuracy of area estimation for both methods. The achieved accuracy of the neural network's classification was 94%. For images without a visible heart area, the subtraction of a heart template proved to be a pragmatic approach with good results.Significance.In summary, we developed a routine for estimation of the lung area combined with estimation of the heart area in electrical impedance tomography images.