An Innovative Learning-by-Examples Method for Real-Time Electrical Impedance Tomography of the Human Chest

A non-invasive diagnosis method for real-time human chest monitoring is presented. The developed technique is based on an innovative learning by examples (LBE) methodology to efficiently solve the electrical impedance tomography (EIT) inverse problem. Towards this end, the partial least squares (PLS) feature extraction is profitably combined with an output-space-filling (OSF) adaptive sampling to build a low-size but highly-informative training database, which is successively exploited to make accurate real-time predictions of the lungs conductivity through support vector regression (SVR). A preliminary numerical validation is shown to assess the potentialities of the proposed LBE inversion technique.