Characterizing upper airway occlusion during natural sleep could be instrumental for studying the dynamics of sleep apnea and designing an individualized treatment plan. In recent years, obstructive sleep apnea (OSA) phenotyping has gained attention to classify OSA patients into relevant therapeutic categories. Electrical impedance tomography (EIT) has been lately suggested as a technique for noninvasive continuous monitoring of the upper airway during natural sleep. In this paper, we developed the automatic data processing and feature extract methods to handle acquired EIT data for several hours. Removing ventilation and blood flow artifacts, EIT images were reconstructed to visualize how the upper airway collapsed and reopened during the respiratory event. From the time series of reconstructed EIT images, we extracted the upper airway closure signal providing quantitative information about how much the upper airway was closed during collapse and reopening. Features of the upper airway dynamics were defined from the extracted upper airway closure signal and statistical analyses of ten OSA patients' data were conducted. The results showed the feasibility of the new method to describe the upper airway dynamics during sleep apnea, which could be a new step towards OSA phenotyping and treatment planning.