Ventilation/perfusion mismatch measured by electrical impedance tomography at the bedside: potentialities and challenges

Ventilation–perfusion (V′/Q) derangements are ubiquitous in critical care, from non-perfused lung regions in pulmonary embolism to the marked V′/Q heterogeneity of acute respiratory failure, where shunt predominant [1, 2]. While the multiple inert gas elimination technique (MIGET) remains the physiological gold standard [3], it only offers global information, and is relatively complex to implement. The prospect of bedside, regional V′/Q measurements using electrical impedance tomography (EIT) has therefore drawn considerable interest.

Obtaining V′/Q images

While both scintigraphy and computed tomography can provide ventilation/perfusion analyses, they require patient transport and imply ionizing radiations. EIT is a bedside, radiation-free alternative extensively validated [4, 5]. All contrast agents carry minor intrinsic risks: ~ 5–10% hypertonic saline is becoming standard in patients [6]. For ventilation, a few breaths are acquired and analyzed according to standard methodology [6], then, for perfusion, apnea is required to inject and properly track the saline bolus (Fig. 1a–d). In intubated patients, this is obtained by ventilator occlusion. Spontaneously breathing patients, instead, should voluntarily hold their breath [6]. A few seconds is waited for stabilization and to estimate the apnea-related drift in impedance [4], then ~ 10 ml of contrast agent is pushed through a central line in < 1–2 s. Apnea normally lasts ≥ 8 s [6], but longer times have been used experimentally and may sometimes be advisable.

Fig. 1.

Schematic workflow for obtaining V′/Q matching by EIT. a after obtaining a few undisturbed tidal breaths (blue rectangle), apnea is induced and the hypertonic saline bolus is injected, resulting in a negative deflection visible on conventional EIT monitors (red rectangle). Raw traces (black dashed line) are low-pass filtered (black continuous line) to remove cardiogenic oscillations. A downward drift in impedance is visible during apnea (purple dashed line). b the inverted impedance trace caused by the hypertonic saline passing through the pulmonary circulation is shown. The global trace is thought to result from summation of three compartments (see annotations). Perfusion is estimated as the maximum upstroke of the lung compartment (red dashed line). c impedance variations caused by tidal ventilation. d the pixel-by-pixel amplitude of tidal impedance variations results in ventilation images. e the pixel-by pixel maximum upstroke of the lung compartment is used as perfusion. f the ventilation and perfusion maps may be superimposed pixel-by-pixel after thresholding, obtaining regional V′/Q ratio maps. g the distribution of ventilation or perfusion across the V′/Q ratio range can be obtained in a MIGET-like fashion, with the addition that the displayed graph can be drawn for any given region of the EIT image. a.u. arbitrary units, EIT electrical impedance tomography, MIGET multiple inert gas elimination technique, t time, t_A arrival time of the saline bolus, V′/Q ventilation/perfusion ratio, Z impedance

Three components are typically combined in most thoracic perfusion traces, termed [4, 5] pre-lung—likely corresponding to the right heart and greater pulmonary vessels—lung and post-lung—encompassing an early left heart rebound, bronchial circulation and recirculation [5] (Fig. 1b). Approximate peak times for each component depend on hemodynamics [7] and have been reported in swine [5, 7], while only inferred from physiology in humans. The low spatial resolution of EIT magnifies the partial volume effect (PVE), which must be addressed by post-processing. Several algorithms are in use to isolate and reconstruct lung perfusion images, greatly differing in how they deal with PVE. The best validated [4, 5, 7] fit multiple gamma variates to data, only to retain the lung signal. Others remove a cluster of cardiac pixels or manually select a time frame between the pre-lung and lung peaks.

Irrespective of the algorithm, regional perfusion is assumed to be proportional to the maximum upslope of the indicator signal, disregarding wash-out (Fig. 1b, e) [5].

To date, the need for stable apnea warrants caution when analyzing indicator-based V′/Q by EIT in spontaneously breathing patients [6, 8]. Alternatively, the pulmonary pulse can be isolated from EIT. However, while related to stroke volume passing from the heart to lung vessels, pulsatility cannot be equated to perfusion, especially as its amplitude appears to increase in collapsed lungs [4, 5]. Rather than a proxy for perfusion, it may be regarded as a promising complement.

Physiological indices of V′/Q mismatch

Most indices rely on the fraction of non-perfused and/or non-ventilated (i.e., "unmatched") pixels. Simple and validated [2], this approach does not require measurements of minute ventilation and cardiac output, but relies on patient-specific thresholds, often reflecting heterogeneity more than actual V′/Q (Online Resource 1, Fig. S1).

V′/Q mismatch may be more precisely quantified. Typically, five compartments are identified (Fig. 1f), equally spaced on a log scale [9], both at the global and regional level: dead space, high, normal, and low V′/Q, shunt. This requires precise measurements of minute volume and cardiac output though non-invasive alternatives are being researched [10], but their units are absolute V′/Q ratios, rather than fractions of pixels.

Despite already being advanced, the compartments can be made finer (e.g., 21), appearing similar to a regional MIGET(Fig. 1g) [9, 10], and descriptive statistics can be calculated, such as mean and standard deviation, potentially less affected by arbitrary selection of V′/Q thresholds.

Limitations of EIT-derived V′/Q

Pixel-by-pixel superimposition (Fig. 1f) of ventilation and perfusion has been questioned as known spatial biases result in perfusion being shifted ventrally [5]. Pendelluft (i.e., intra-pulmonary airflows) may also complicate the picture.

An arbitrary threshold of 5–20% of the maximum value within the image is often employed for the need to remove noise to non-perfused and/or non-ventilated pixels [2]. This defines unmatched pixels and heavily influences the extreme compartments (i.e., dead space and shunt) of the other approaches [10].

EIT can track perfusion changes, but exhibits poor baseline agreement with conventional imaging [4]. The same applies to EIT-based V′/Q when compared to capnography and blood gasses [2]. However, regional information is unique to EIT.

Finally, EIT only images roughly half of the chest, so that changes due to PEEP and body position should be interpreted cautiously, specifically in patients with more heterogenous non-focal disease.

Clinical significance of EIT-based V′/Q

Experts broadly agree that EIT can quantitatively evaluate the effects of body positions on V′/Q [6]. In the prone position, an early dorsal redistribution of ventilation occurs, and a paradoxical anti-gravitational redistribution of perfusion follows after the first 3 h leading to better V′/Q matching [11]. Interestingly, slightly different kinetics exist, according to patient phenotypes [12]. However, measurements of perfusion by EIT in the prone position may still require methodological refinements [6].

Higher PEEP consistently improved V′/Q matching [9] by decreasing shunt and—surprisingly—dead space in patients with higher recruitability. On the other hand, when over-distension predominates at higher PEEP, V′/Q matching could worsen, so that personalized PEEP by EIT balancing over-distension and collapse could be key to optimal V′/Q matching [13].

Finally, EIT could demonstrate that the oxygenation response to inhaled nitric oxide is associated with ventral redistribution of lung perfusion [14].

Unmatched units predicted lower survival in a cohort of ARDS patients, apparently outperforming P/F and ventilatory ratio and independent from SAPSII score [2]. Other V′/Q mismatch indices were also related to clinical outcomes [10], could identify focal/non-focal ARDS at the bedside and could support the diagnosis of pulmonary embolism with rather high specificity [15].

Conclusions

V′/Q mismatch measured by EIT is steadily evolving into a bedside tool for personalized patient care. Current evidence supports its use as an adjunct to established techniques for the initial evaluation of pulmonary embolism and respiratory failure and for monitoring patient responses to ventilatory maneuvers [6].

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open access funding provided by Università degli Studi di Milano within the CRUI-CARE Agreement. Ministero della Salute, Current research from Ministry of health, Tommaso Mauri, Rome, Tommaso Mauri, Italy, Tommaso Mauri.

Data availability statement

There are no data.

Declarations

Conflict of interest

TM received personal fees for speaking at sponsored symposia by Draeger, Fisher and Paykel, and Aerogen unrelated to the present work. All other authors, none.