ELECTRONIC SUPPLEMENTARY MATERIAL
Evaluation of an electrical impedance tomography-based global inhomogeneity index for pulmonary ventilation distribution
Zhanqi Zhao1,2, Knut Möller2, Daniel Steinmann1, Inéz Frerichs3, Josef Guttmann1
1Section for Experimental Anesthesiology, Department of Anesthesiology and Critical Care Medicine, University Medical Center, Freiburg, Germany
2Department of Biomedical Engineering, Furtwangen University, Villingen-Schwenningen, Germany
3Department of Anesthesiology and Intensive Care Medicine, University Medical Center of Schleswig-Holstein Campus Kiel, Germany
Address of the corresponding author:
Zhao, Zhanqi MSc.
Anaesthesiologische Universitätsklinik Freiburg
Sektion für Experimentelle Anaesthesiologie
Hugstetter Straße 55, D-79106 Freiburg, Germany
Tel: +49-761-2702309
Fax: +49-761-2702328
The identification of the lung area based on EIT:
The Lung Area Estimation (LAE) method
Hahn and Frerichs et al. have proposed a functional EIT (fEIT) based identification of the lung boundaries in EIT images. They assume that a pixel belongs to the lung area if the variation of the pixel values in the EIT images during a certain period is larger than a threshold of 20% of the maximum variation [S1-S5]. This approach, called fEIT method in the following, may fail to exclude the heart from the lung. Further in some lung diseases, such as lung cancer and pneumothorax, the ventilation in parts of the lung region may be impaired to such an extent that it is not visible at all in a fEIT image.
We recently proposed a novel method for lung area estimation (LAE) [S6]. The LAE method can be described in 3 steps:
1. Identify the lung area with the fEIT method.
2. Mirror the lung area identified in the first step from left to right and from right to left and combine them by a logical OR-operation. The overlaid lung areas of both sides are projected back to the other side.
3. Subtract the cardiac related area from the lung area derived in the second step. Since the heart rate is normally different from respiratory frequency in adults, these two kinds of areas may be well separated in the frequency domain. The impedance variation of every pixel vs. time was analyzed in the frequency domain by Fast Fourier transform (FFT). The total power in the frequency range below (respiratory value) and above 0.8 of the heart frequency (heart value) was determined. The ratio between these two values of power was then compared to an individual threshold θ (Eq. S1). Pixel ratios smaller than the threshold were subtracted from the lung area.
(S1)
where Ek(f) represents the energy distribution function of pixel k in frequency domain. hf corresponds to the heart frequency in Hz, sf denotes the sampling frequency in Hz. θ is the threshold determined individually. Let
(S2)
and RE, 1:n is a vector of all the RE sorted in ascending order of its value. Fig. S1 shows RE,1:n of one patient. θ is equal to the first RE value in RE, 1:n satisfied RE, i+1 − RE, i < 0.0014, i є [1,n).
As a result, a quasi-symmetric left and right lobe results that includes all detectable lung but not cardiac related area.
There are two types of fEIT scans [S7, S8]. The first type calculates the standard deviation of relative impedance change in each image pixel for a certain time period [S9]. The second type calculates the linear regression coefficient in the following regression formula:
(S3)
where Local ΔZ and Global ΔZ are local (each pixel) and global (sum of all pixels) relative impedance change for a certain time period, respectively; α, β are regression coefficients and ε is the fitting error [S10]. The standard deviation values in the first type or the regression coefficient α in the second type are plotted in the corresponding pixel locations to build a fEIT image. It is suggested that the second type of fEIT should be used when the environment during data acquisition is very noisy [S7, S8, S11]. But the “noise” that the second type fEIT suppressed may help to assess the pulmonary perfusion [S12], which is essential to identify the non-ventilated lung area (e.g. atelectasis). In particular, the noise level in the present patient data was low and these two types of fEIT images were similar (Fig. S2). Considering the fact that the first type of fEIT may have potential benefits for the collapsed lungs, it was used for LAE method and for fEIT method in stead of the second type.
Comparison of LAE method and fEIT method
Totally 50 patients were studied and divided into control group and test group. For details of patient’s information and protocols, please refer to the “Materials and methods” section of the original article. EIT measurement was performed and the EIT images were analyzed.
Fig. S3 shows the lung area determined for a patient in control group with fEIT method and LAE method, respectively. Cardiac related area was excluded in the result of LAE method. Fig. S4 shows the lung area identified with these two methods for a patient in the test group. Part of the lung was collapsed and thus was not detected by the fEIT method.
It is assumed that the fraction of the lung area in the thorax cross-section for different people should be more or less in the same range, in spite of the state of the lung. With fEIT method, the size of lung area identified in the control group SC-fEIT = 361 ± 35.1 (mean ± standard deviation) and in test group ST-fEIT = 299 ± 60.8. They differed significantly (P = 0.004). On the contrary, the sizes estimated with LAE method in control group SC-LAE = 353 ± 27.2 and in test group ST-LAE = 353 ± 61.1. They did not differ significantly (P = 0.41). Figure S5 shows the comparison of the size of the identified lung area in quartiles.
The special ventilator settings in the control group may have influenced the results because zero PEEP together with the high FIO2 during induction of anesthesia is expected to lead to a derecruitment. In the worst case, large parts of dependent lung regions in patients from the control group were collapsed. Due to the mirroring, the lung area determined by the LAE method in the dependent lung regions will be greater or equal to the atelectic lung area determined with the fEIT method, i.e. Sdependent,C-LAE ≥ Sdependent,C-fEIT. If both LAE and fEIT methods are compared with cardiac related areas being subtracted, the mirroring effect alone adds on the average 7% more pixels to the results of the LAE method. Since the original fEIT method includes also cardiac related area, small reduction of the whole lung area size SC-LAE compared to SC-fEIT (Fig. S5) may be due to the subtraction of the cardiac related area by the LAE method. The enlargement of ST-LAE compared to ST-fEIT may be mainly due to the invisibility of the damaged part in one lung by the fEIT method and again the mirror effect of the LAE method, which leads to the increase of P from 0.004 (between control and test group, with fEIT method) to 0.41 (with LAE method). However, it has also been observed that although the LAE method was used, the estimated size of the lung in the test group is still a little bit smaller than in the control group (Fig. S5). The lungs may be damaged to such an extent that neither of both lungs is completely visible in the EIT image. That leads to a decrease of GI and makes the air distribution seem more homogenous. Even so the GI is still high compared to the control group (Fig. 2 in the original article). Further anatomical confirmation with computed tomography (CT) or single photon emission CT (SPECT) is needed.
Comparison of LAE method and VT method
Hahn et al. have recently proposed an improved approach (VT method) to image ventilation in fEIT [S8]. In principle it could be also used to determine the lung area by adding a threshold value. Such a lung area determination method will be called VT method in the following. A direct comparison of the LAE method and the VT method is not possible since the calculation of the VT method is based on raw data, which is not available to us. But according to the description of the authors, the VT method is superior to the first type of fEIT by suppressing the non-ventilatory impedance changes (noise and impedance changes, e.g., from heart action and movement artifacts). Unfortunately, both VT method and fEIT method may fail to discover the non-ventilated part of the lung (e.g. atelectasis). On the other hand, the LAE method tries to identify the non-ventilated lung area by mirroring the ventilated lung area.
Actually, both VT method and fEIT method can be used as the first step of the LAE method. However, as mentioned before, excluding the cardiac related area at the beginning (the second type of fEIT and VT method) may fail to assess lung aeration therefore fail to percept the non-ventilated lung area. Besides, the phase shift because of the inhomogeneous air distribution in the lung may influence the VT method but not the fEIT method. Nevertheless, using the fEIT method may include different source of noise other than cardiac related actions. Considering the noise level in the present patient data was low, the LAE method was calculated based on the fEIT method.
Limitations of LAE method
Heart may not be the only factor that influences the symmetry of the left and right lungs (e.g. the mediastinum). By mirroring, some pixels which do not belong to the lung area are misleadingly added to the lung area. Subtraction of the cardiac area may not be enough but nevertheless the error is expected to be minimized.
The LAE method may not perceive the lung regions when both lungs collapse at the same corresponding areas. Multi-frequency EIT may perform better in such extreme case since large impedance differences may exist in different tissues [S13]. The quality and interpretation of multi-frequency EIT measurements is still difficult and further investigations are needed.
REFERENCES
S1. Frerichs I, Hinz J, Herrmann P, Weisser G, Hahn G, Dudykevych T, Quintel M, Hellige G (2002) Detection of local lung air content by electrical impedance tomography compared with electron beam CT. J Appl Physiol 93: 660-666
S2. Frerichs I, Hahn G, Schroder T, Hellige G (1998) Electrical impedance tomography in monitoring experimental lung injury. Intensive Care Med 24: 829-836
S3. Frerichs I, Hahn G, Golisch W, Kurpitz M, Burchardi H, Hellige G (1998) Monitoring perioperative changes in distribution of pulmonary ventilation by functional electrical impedance tomography. Acta Anaesthesiol Scand 42: 721-726
S4. Frerichs I, Hahn G, Hellige G (1999) Thoracic electrical impedance tomographic measurements during volume controlled ventilation-effects of tidal volume and positive end-expiratory pressure. IEEE Trans Med Imaging 18: 764-773
S5. Hahn G, Frerichs I, Kleyer M, Hellige G, (1996) Local mechanics of the lung tissue determined by functional EIT. Physiological measurement 17 Suppl 4A: A159-166
S6. Zhao Z, Möller K, Steinmann D, Guttmann J (2009) Determination of lung area in EIT images. In: Proceedings of the 3rd conference on iCBBE, Beijing, China. IEEE (in press)
S7. Pulletz S, van Genderingen HR, Schmitz G, Zick G, Schadler D, Scholz J, Weiler N, Frerichs I (2006) Comparison of different methods to define regions of interest for evaluation of regional lung ventilation by EIT. Physiol Meas 27: S115-127
S8. Hahn G, Dittmar J, Just A, Hellige G (2008) Improvements in the image quality of ventilatory tomograms by electrical impedance tomography. Physiol Meas 29: S51-61
S9. Hahn G, Sipinkova I, Baisch F, Hellige G (1995) Changes in the thoracic impedance distribution under different ventilatory conditions. Physiol Meas 16: A161-173
S10. Kühnel G, Hahn G, Frerichs I, Schröder T, Hellige G (1997) Neue Verfahren zur Verbesserung der Abbildungsqualität bei funktionellen EIT-Tomogrammen der Lunge. Biomed Technol Berl 42 Suppl: 470-471
S11. Pulletz S, Elke G, Zick G, Schadler D, Scholz J, Weiler N, Frerichs I (2008) Performance of electrical impedance tomography in detecting regional tidal volumes during one-lung ventilation. Acta Anaesthesiol Scand 52:1131-1139
S12. Schibler A, Calzia E (2008) Electrical impedance tomography: a future item on the "Christmas Wish List" of the intensivist? Intensive Care Med 34: 400-401
S13. Oh TI, Koo H, Lee KH, Kim SM, Lee J, Kim SW, Seo JK, Woo EJ (2008) Validation of a multi-frequency electrical impedance tomography (mfEIT) system KHU Mark1: impedance spectroscopy and time-difference imaging. Physiol Meas 29: 295-307
FIGURE LEGENDS
Figure S1: RE, Ratio of energy distributions between two frequency ranges. RE were sorted in ascending order. The higher the RE was, the higher the respiratory related activity presented. N is the number of pixels. Dashed line denotes the threshold θ in Eq. S1. Figure adapted from Ref. S6
Figure S2: Two different types of fEIT images. Left, the first type: the standard deviations of relative impedance change in each image pixel for a certain time period were used; right, the second type: the regression coefficient α in Eq. S3 were used.
Figure S3: Comparison of the lung area identified with different methods for a patient in control group. A: fEIT image of the patient; B: fEIT method and C: LAE method.