3. HIPERELASTICIDAD ANIS ´ OTROPA
3.2. Modelo is´otropo de Sussman y Bathe
Our findings indicate that the trending of SV by means of EIT is not possible based on the heart amplitude. This is not only in contradiction with the simulation-based findings presented in Chapter 6 but also partly with the previous studies by Vonk Noordegraaf et al. [155] and Pikkemaat et al. [114]. We hypothesize that in our case the heart-based SV estimation is impaired by ventilation-induced out-of-EIT-plane movement of the heart and/or changes in the heart-lung conductivity contrast (see Figure 6.9) caused by the injection of fluid or other influences. None of these effects was considered in simulations which might explain the contradictory results. While in the pig experiments by Pikkemaat et al. more accurate belt placements might have lead to a less pronounced influence of the first effect in certain pigs, it was mentioned as a main limitation for other pigs [114].
In contrast, SV trending via the lung amplitude is more promising, which would confirm the recent work by da Silva Ramos et al. [41] but is at the same time partly in contradiction with the mixed results obtained by Pikkemaat [113]. Besides, these two studies [113, 41] were performed on pigs under laboratory conditions while our study is based on clinical data of ICU patients under real-life conditions.
9.3.6 Limitations and Future Work
Although we performed measurements on 16 patients, most of them did not show large SV variations in response to the fluid challenge, i.e. not many SV changes exceed the±13 % error expected from thermodilution. Therefore, in a future study more patients and in particular those with higher SV variations should be included.
While our results suggest that the EIT lung amplitude is able to track changes in SV and thus global perfusion, it is of utmost importance to underline the following. There is still the widespread assumption that cardiosynchronous EIT signals in the lung region mainly reflect pulmonary perfusion which is not entirely correct as discussed by Hellige and Hahn [74] and Adler et al. [11], i.e. two pulmonary arteries with equal perfusion but different compliance result in different pulsatile EIT signal amplitudes – despite equal perfusion! Even though the present approach shows promise on the current data, further research – including long- term measurements and various pathophysiological conditions – is required to reveal likely limitations of this approach.
The current EIT images are based on one single – patient-independent – reconstruction model. It is known that a mismatch between actual thorax shape and reconstruction model can impact the measurements [64]. Therefore, it would be interesting to investigate to which extent the measurements are improved when considering a patient-specific morphology for EIT reconstruction.
9.4 Conclusion
In view of noninvasive SV monitoring at the bedside we performed EIT measurements on six- teen patients in the ICU. These measurements were performed before and after fluid challenge which is known to increase the SV in those patients which are fluid responsive. Resulting chan- ges in EIT-derived heart and lung amplitudes were then compared to SV changes obtained via transpulmonary thermodilution.
Our analysis suggests that the trending of SV – i.e. following relative changes – by means of EIT is not possible based on the heart amplitude, which is in contradiction with previous studies by Vonk Noordegraaf et al. [155] and Pikkemaat et al. [114]. In contrast, SV trending via the EIT-derived lung amplitude is more promising, but only after excluding 9 out of 32 (28 %) measurements considered as potential outliers with too high noise level. In this case an acceptable trending performance with CR=100 % and²α= −1.6±11.9° is obtained. When compared to pig experiments, our findings would confirm the recent work by da Silva Ramos et al. [41] but are partly in contradiction with the mixed results obtained by Pikkemaat [113]. As these promising results are based on solely 8 measurements of 5 patients, further research is required to validate this approach on more patients and in particular with higher variations in SV. Moreover, it is known that the EIT lung amplitude is not merely related to pulmonary perfusion [74, 11]. Therefore, this potential limitation needs to be further investigated in dedicated clinical trials including long term measurements and addressing relevant pathop- hysiological conditions. Nonetheless, continuous and noninvasive SV estimation might be feasible in environments such as the ICU where controlled EIT measurements can be perfor- med. Moreover, absolute SV estimates could be obtained by scaling the relative changes with an initial calibration value (e.g. obtained via transthoracic echocardiography).
Part IV
Towards an Optimized
Measurement Setup for
EIT-Based SV Monitoring
10
Considerations for an Improved Mea-
surement Setup
In this chapter we briefly review existing limitations of available EIT systems and discuss potential approaches for finding a measurement setup better suited for EIT-based SV monito- ring.
10.1 Practical Limitations of Available Clinical EIT Systems
The EIT systems currently available and certified for monitoring in clinical scenarios (see Table 3.1) have certain limitations when aiming for SV monitoring via the heart amplitude. First, most of these systems have the EIT electrodes embedded in a belt which restricts the image reconstruction onto one single plane (2D). Yet, as shown in Chapter 6, the EIT-based SV estimates are considerably influenced by the level at which the EIT belt is placed. This is also indicated by the experimental measurements in the OR (see Chapter 8) where the data from two patients had to be excluded from analysis due to a too low heart signal (presumably caused by a too low/high EIT belt placement). Besides, even if the EIT belt was systematically placed at the same level, the EIT heart signal can be altered by respiration- or posture-induced (out-of-EIT-plane) heart displacement [170]. It is therefore assumed that the use of 3D EIT would be of great benefit to minimize these undesired influences.
Second, all of the available clinical EIT systems are limited to a bipolar stimulation and measurement pattern with one specific skip (i.e. number of inactive electrodes in between the two ones actively measuring voltage/injecting current). Moreover, certain systems (e.g. Göttingen Goe-MF II and Dräger PulmoVista™ 500) make use of the adjacent (skip=0) pattern, which is known to have the lowest performance in terms of sensitivity and signal-to- noise ratio (in the center) [134, 8]. While the other systems use a skip>0, to the best of our knowledge, they do not provide a possibility to change it. Even though different skips can theoretically be achieved by connecting the electrodes (of a skip 0 system) in a different order [8], this is not always practicable as the resulting increase in raw voltage amplitudes leads to saturation of the analog acquisition chain.
?
e1 e2 e3 e4 e5 e6 e32 e7 c1 c2 c3 c4 c5 c6 c32 c7 EIT Device ... ... 1 4 5 8 32 29 28 2 3 6 7 31 30 27 (a)?
e1 e2 e3 e4 e5 e6 e32 e7 c1 c2 c3 c4 c5 c6 c32 c7 EIT Device ... ... 1 4 5 8 32 29 28 2 3 6 7 31 30 27 (b)Figure 10.1 – Two electrode placements assumed to be more suitable for EIT-based SV mo- nitoring: (a) a patch of electrodes close to the heart, and (b) two transversal planes of 2×16 electrodes. See text in Section 10.2.
Thirdly, it is known from measurements on healthy subjects that the strength of cardiosyn- chronous EIT signals decreases with increasing stimulation frequency [30]. On the other hand, due to electrical safety considerations, the injected current and thus the signal-to-noise ratio is lower at frequencies below 100 kHz [2]. However, only some devices allow changing this frequency (e.g. Göttingen Goe MF II and Dräger PulmoVista™ 500) while others are fixed to a given frequency (e.g. at present Swisstom BB2is fixed to 195 kHz).