Measuring regional lung ventilation/perfusion ratio using electrical impedance tomography

点击翻译按钮获取中文摘要

For the treatment of intensive care patients with acute respiratory distress syndrome (ARDS) careful guidance of mechanical ventilation is essential. Ventilation therapy must ensure alveolar gas exchange and at the same time avoid any further ventilator-associated lung damage. Electrical impedance tomography (EIT) is a noninvasive imaging modality that allows real-time monitoring of regional lung ventilation (V) at bedside with high time resolution. However, since optimal alveolar gas exchange only takes place in lung regions that are both adequately ventilated and perfused, it would be highly beneficial to also obtain regional lung perfusion (Q) and regional ventilation to perfusion ratio (V/Q) using EIT. In the previous project, a model-based approach to determine Q using EIT with contrast-agents (indicator-based signal, IBS) was developed and validated in combination with V against single-photon emission computed tomography (SPECT) under physiological condition. In addition, extensive animal studies were performed on the suitability of low-dose contrast agents for use in patients. The proposed follow-up project focuses on the determination of V/Q using the IBS. In addition, it will be investigated whether it is possible to determine Q from the pulsatile, cardiac-related signal (CRS) of EIT. In an animal study, physiological and pathological V/Q ratios from EIT will be validated against SPECT. At the same time, a coupled simulation model will be developed to comprehensively investigate the origin of IBS. Based on experimental and simulation data, the model-based approach will be adapted to pathological conditions. In order to bring the method closer to clinical application, it will then be tested whether it is also possible to obtain Q from the IBS without interruption of mechanical ventilation. In order to obtain V/Q, it will be investigated whether EIT reconstruction can be optimized for problem specific solutions using machine learning (Convolutive Neural Networks, CNN). Finally, it will be examined whether Q from the CRS can be improved using the IBS as a reference.