After having isolated the cardiosynchronous changes, the resulting EIT image sequences can be processed in view of hemodynamic parameter estimation (see Step 4 in Figure 3.1). In the current section we review previous work with particular focus on EIT-based SV estimation. Compared to lung-related EIT, only little work has been published on the topic of cardiovascu- lar EIT. Most of the studies targeted regional pulmonary perfusion which can be assessed using injections of hypertonic saline as contrast agent [18, 50] or other – more indirect – approaches listed in [52]. Besides this, Solà et al. [137] reported on the feasibility of estimating central blood pressure via EIT using the pulse wave velocity principle in the descending aorta. Two older studies address the signal interpretation [46] or suggest a better belt placement [154] for cardiac imaging. Very recently, Proença et al. [120] have successfully shown the measurement of pulmonary artery pressure via the pulse wave velocity assessed in the lung region. Apart from that, most of the remaining literature in cardiovascular EIT concentrated on the estima- tion of SV or CO via EIT which is presented hereafter in more detail.
(a) In 2000 Vonk Noordegraaf et al. [155] were the first to report on the estimation of SV via EIT. Using pulmonary artery catheter thermodilution as a reference they trained their algorithm on 23 patients. The calculations are based on the impedance change from a region hypothesized to stem solely from ventricular activity. However, they could not find a direct correlation between SV and this impedance change. Without further reasoning they included a timing parameter as a second variable into their model to achieve a better fit. The algorithm was then successfully validated on 11 healthy volunteers showing an error of 0.7±5.4 mL, when compared to cardiac MRI as SV reference. In contrast to the traditional transversal EIT belt placement, they placed the electrodes in an oblique plane, which is expected to improve the separation of atrial and ventricular activity in the EIT images [154]. Data was measured using the Sheffield mark I system [27] and processed after ECG-gating of 200 cardiac cycles.
(b) Some years later, in 2006, Zlochiver et al. [173] suggested to avoid the generation of images by using a parametric reconstruction approach. Hence, an ellipsis representing
3.2. Previous Work on EIT-Based Stroke Volume Estimation
the left ventricle was optimized to best fit the impedance values measured. The SV was then simply estimated as the changes in volume of this ellipsis. A major drawback of this study is the use of the controversial ICG technique (see Section 2.2.3) as reference measure for SV. Furthermore, the approach is highly dependent on anatomical a priori knowledge. In a later publication the same group presented a theoretical study – again by using parametric EIT – showing the feasibility of SV estimation via six internal electrodes of an implanted pacemaker device [102].
(c) In 2014 Pikkemaat et al. [114] demonstrated the feasibility of estimating SV in 14 pigs via the heart-related impedance change by using a subject-specific one-point calibra- tion. In certain animals they observed an unclear scaling of the heart amplitude and hypothesized that it might be related to lung volume and also to the relative position of the heart with respect to the EIT electrode plane. Due to these unresolved issues they underlined the need for further investigations. Moreover, they questioned the accuracy of thermodilution PCA as their SV reference and suggested using a second measurement technique for validation purposes. Data was measured using the Dräger EIT Evaluation Kit 2 and processed after PCA-based decomposition [112, 113].
While not available in the corresponding publication [114], in his thesis [113], Pikkemaat also reports on the analysis of the lung-related impedance change zSVp. In experi-
ments where SV was modulated by changing the ventilation settings (the positive end- expiratory pressure – PEEP),zSVpshowed a higher correlation with SV (r=0.69) when
compared to the heart amplitudezSVc (r =0.64). On the other hand, when SV was
modified using Dobutamine – a positive inotropic agent –zSVpcorrelated less with SV
(r=0.61) thanzSVc(r=0.64).
(d) Very recently da Silva Ramos et al. [41] performed investigations in twelve pigs where EIT-based SV was estimated from the systolic amplitude in the lung region (∆Zsys). Large
variations in SV were induced via hemorrhagic shock and subsequent fluid challenges.
∆Zsyswas compared to SV reference measurements from transpulmonary thermodilu-
tion and showed an acceptable trending ability (91.2 % concordance rate). In contrast, absolute SV measurements were only accurate when taking into account body dimen- sions. Data was measured using the Timpel ENLIGHT™ device and processed after ECG-gating with a fixed time window of unknown length.
(e) Maisch et al. [96] were not estimating SV itself but its variations induced by ventilation (known as SVV) and showed the feasibility of measuring this quantity in eight pigs by means of EIT. Measuring SVV is of interest since it can be used to assess fluid re- sponsiveness and therefore help to improve the intravascular volume of mechanically ventilated patients [99]. Similar to the aforementioned method of central blood pressure estimation [137], this approach exploits signal information from the descending aorta. In 2017 Trepte et al. [148] reported on further investigations on 30 other animals where EIT-based SVV estimation showed to be feasible in healthy lungs but comprised in acute lung injury. In both studies data was measured using the Timpel ENLIGHT™ device.
All the studies presented above show the necessity for further investigations and validations, in particular for measurements on humans in real clinical environments – the main goal of this thesis.