Watershed functional lung space: a pendelluft-aware EIT segmentation method

Objective. Pendelluft is the movement of air between lung regions, which results in an apparent phase-shift of the aeration curve between lung regions. Electrical impedance tomography (EIT) can be used to detect and quantify pendelluft at the bedside. A common method to select the functional lung space (FLS)-i.e. the pixels associated with ventilation-is applying a threshold to the pixel tidal impedance variation (TIV). Due to the apparent phase shift, pixel TIV is lower in regions associated with pendelluft, resulting in removal of those pixels for further analysis.Approach. We developed a novel method for FLS selection using the established watershed segmentation method. Watershed regions are segmented based on the pixel amplitude map and local peaks in that map. Watershed regions whose local peaks fall inside the threshold-based TIV FLS are included. Pixels with an amplitude above the threshold that are inside the included watershed regions form the Watershed FLS.Main results. We evaluated the algorithm in 11 patients switching from controlled mechanical ventilation (CMV) to assisted mechanical ventilation (AMV). No significant differences were found between TIV FLS and Watershed FLS during CMV. Switching from CMV to AMV lead to a significant decrease of the TIV FLS (p= 0.043), but not the Watershed FLS. As a result, TIV FLS was significantly smaller than Watershed FLS during AMV (p⩽ 0.001). Pendelluft magnitude was higher using the Watershed FLS compared to the TIV FLS during AMV (p⩽ 0.001).Significance. The common TIV-based FLS can result in the unintended removal of pendelluft-associated pixels for further analysis. The Watershed FLS includes these pixels, potentially improving the quality of EIT-analysis in patients with spontaneous breathing efforts.