Automatic Identification of Inflection Points in Pressure and Regional EIT curves

Abstract

Ventilator Induced Lung Injury (VILI) is a serious condition caused by sub-optimal settings of mechanical ventilation in Acute Lung Injury (ALI) patients. The main contributors to VILI are 1) cyclic opening and closing of collapsed lung tissue which occur at low pressure and 2) overdistension of lung tissue which occur at high pressures. Reducing these within mechanically ventilated patients can lead to an increase in likelihood of survival. The key clinical measure to reduce VILI is selecting an appropriate Positive-End Expiratory Pressure (PEEP) to make a balance between keeping lung units open while not overdistending them. Electrical Impedance Tomography (EIT) provides regional lung air volume information which promises to help improve clinical selection of PEEP. The goal of this paper is to compare two automated methods (three-piece linear and sigmoid models) and one manual method (visual heuristics) in the analyse of EIT data to locate regional inflection points (IP). These IP can be used to distinguish between collapsed and overdistended regions, thus assisting in the location of a best suited PEEP. These algorithms were implemented, tested, and compared to previously suggested approaches, using a clinical database of ALI and healthy lung patients. Results varied depending on which IP was being compared. Comparing the visual heuristic method with the linear spline method differences ranged from -1.507 mbar to 0.0240 mbar. The results are promising and continued work on the linear method in IP selection is suggested.