While often a live-saving therapy and a necessity during general anesthesia, mechanical ventilation may promote lung damage when ill-adapted, a phenomenon called ventilator-induced lung injury (VILI) (1). Some patient populations have been shown to be at higher risk of developing VILI: patients suffering from the acute respiratory distress syndrome (2) and patients requiring one-lung ventilation (3).

Assessing mechanical properties of the lung has been suggested as a promising approach to detect conditions leading to VILI (4) with excessive pulmonary strain thought to be a critical factor in its development (5, 6). Unfortunately, pulmonary strain is non-trivial to measure at the bedside (7) and most techniques measure global pulmonary strain, an oversimplification of the complex lung physiology. Areas of higher regional pulmonary strain that cannot be assessed using global measures have been implicated in the development of pulmonary inflammation (8). Detecting these areas of higher regional pulmonary strain may enable identification of patients at higher risk of VILI who have seemingly benign global strain measures and yet may benefit from further strain-lowering interventions (9).

While magnetic resonance imaging (10), positron emission tomography (11) and thoracic computed tomography (9) are currently available in a research setting to measure regional pulmonary strain, these techniques are particularly burdensome, subject patients to the dangers of intra-hospital transport (12), expose patients to ionizing radiations (13) and haven’t been validated against a gold standard nor have they been shown to be associated with clinical outcome. Electrical impedance tomography is an increasingly popular monitoring modality in ventilated patients (14) and has the advantage of being a bedside and radiation-free imaging modality. Unfortunately, electrical impedance tomography cannot currently measure regional pulmonary strain and requires costly dedicated equipment that is currently unavailable in most operating rooms and intensive care units. Consequently, all of the above modalities have major drawbacks that limit their use and validity.

Lung ultrasonography, a bedside, precise, repeatable, easy to learn, and low-cost exam (15), is already used to monitor pulmonary aeration during mechanical ventilation (16). Preliminary works hint that it may be an interesting avenue to measure regional pleural strain as a surrogate for regional pulmonary strain (17, 18). Ultrasound (US) systems, now standard equipment in most operating rooms and intensive care, could make bedside regional pleural strain measurement commonplace. However, prospective feasibility data is lacking.

The objective of this study was to assess the feasibility and reliability (intraobserver, interobserver, and test-retest) of pleural strain measurements by ultrasound elastography. Moreover, we sought to determine if elastography indices of pleural translation, strain and shear would correlate with varying tidal volumes.

This report was redacted following the CONSORT 2010 extension to randomized crossover trials statement (19).

Twelve patients were assessed for eligibility. Two patients were excluded post-enrollment (Supplementary Figure 4): one change in anesthetic plan and one consent withdrawal. Patients’ baseline characteristics are summarized in Table 2.

With our protocol specifying that 72 cineloops would be acquired for each of the 10 patients randomized, 720 elastograms were planned to be computed upon completion of cineloop acquisition. Sixty-two cineloops (8.6%) weren’t subjected to segmentation and tracking: 4 cineloops were not acquired because of a manipulation error, 7 cineloops were incomplete because of a technical issue with the echograph, and 51 cineloops didn’t show sufficient pleura throughout the respiratory cycle. This last problem was only encountered for cineloops acquired over the left 3rd intercostal space at the mid-clavicular line where the heart was mostly seen. Lastly, elastograms were not computed for 6 cineloops because of inadequate tracking of the ROI throughout the respiratory cycle. Elastogram computation was thus performed on 652 (90.6%) cineloops.

Estimates for fixed effects of all 6 models can be found in Table 3. We observed a significant linear increase in 4 elastography parameters with increasing tidal volume: lateral absolute shear, lateral absolute strain, Von Mises strain and lateral absolute translation (Figure 2). With overlapping 95% confidence intervals, no significant elastography parameter was superior to the other in its ability to capture an increase in tidal volume (Figure 3). Amongst them, lateral absolute shear, lateral absolute strain and Von Mises strain were more precisely predicted by tidal volume compared to lateral absolute translation, as shown by higher goodness of fit (Table 3). The first sensitivity analysis excluding suboptimal images yielded similar results (Supplementary Table 1 and Supplementary Figure 5). When side and gravity dependence of measurements were considered, only gravity dependence seemed to be associated with our results (Table 3). Dependent lung zones had initially lower values for all significant strain elastography parameters, but pleural strain measurements in dependent lung zones increased more rapidly with increasing tidal volume although this was only significantly with the Von Mises strain parameter (Table 3 and Figures 2, 3).

Measured intraobserver reliability was excellent for lateral absolute shear, lateral absolute strain and Von Mises strain. Interobserver and test-retest measured reliabilities were found to be moderate to good from Von Mises strain and moderate to excellent for lateral absolute shear and lateral absolute strain (Figure 4 and Supplementary Table 2). In our sensitivity analysis, bootstrapping yielded identical estimates (Supplementary Table 3). Bland-Altman plots, mean bias and 95% limits of agreements can be found in Supplementary Table 4 and in Supplementary Figure 6).

We showed preliminary evidence that computing regional pleural translation, strain and shear components is feasible in over 90% of cineloops. Lateral absolute shear, lateral absolute strain and Von Mises strain were more informative in regards to administered tidal volumes than lateral translation, lateral absolute translation and lateral strain. Measured intraobserver reliability was excellent for all significant parameters while interobserver and test-retest measured reliability were found to be moderate to excellent. This supports our hypothesis that lung ultrasound could be used to measure regional pleural strain in mechanically ventilated patients.

While others have previously described ultrasonographically-measured regional pleural strain in humans (17, 18), this study significantly extends previous work by quantifying the impact of tidal volume on regional pleural strain. Despite the absence of a gold standard precluding any firm conclusion about the utility of each elastography parameter, we described an expected dose-response relationship between various strain measurements and administered tidal volumes.

At first glance, it is somewhat surprising that a 2 mm thick ROI delineated from an image generated by a 12 MHz probe could capture physiological signal from the pleura, a structure only 30–40 microns thick (30). With the topmost part of the ROI aligned with the pleura, the image contained within the ROI is actually mainly composed of reverberation artifacts originating from the pleura and other soft tissue/ultrasound beam interfaces (31). These artifacts are most easily seen when contrasting normal lung sliding in M-mode and their disappearance when a pneumothorax occurs (Supplementary Figure 7). We hypothesize that reverberation artifacts in the ROI convey anatomical information on the pleura thus allowing our algorithm to compute elastography parameters. Another possibility would be that imperfect ROI tracking leads to intercoastal muscle being included in the strain calculation. While it is technically possible to measure intercostal muscle strain (32), we do not believe this is the case with our results as patients were paralyzed for induction of general anesthesia.

One important finding is that lateral strain had a poor dose-response with tidal volume despite a time plot of lateral strain that seemed to track delivered tidal volume in many patients (Figure 1C). Perhaps explaining the poor performance of lateral strain compared to lateral absolute strain, we observed small areas of negative strain during lung inflation (Figure 1A). Supporting this hypothesis, zones of local deflation during inspiration have been described in animals (Supplementary Figure 8) (33, 34). Analogous to how lateral strain and lateral absolute strain are calculated, global pulmonary compliance was correlated with the sum of positive values of local compliance, measured by computed tomography, but not with the sum of positive and negative values of local compliance (34). While the exact causes for this phenomenon are still debated, tidal recruitment with gas redistribution may contribute. Discrepancies between lateral strain and lateral absolute strain could signal tidal recruitment and ongoing atelectrauma.

Our study has some important strengths and limitations. First, our results are physiologically plausible with lateral absolute shear, lateral absolute strain and Von Mises strain increasing with increasing tidal volumes. Second, our two sensitivity analyses demonstrate the robustness of results. Third, by performing a randomized crossover study, we ruled out the possibility of an order effect. Fourth, we demonstrated the good to excellent reliability of these measurements. Given the proof-of-concept nature of this study, several limitations should be considered. First, the limited sample size restricts our ability to obtain precise estimates of pleural strain in human. Second, no gold standard for pulmonary strain was available for comparison. While helium dilution or nitrogen-washout techniques could have been used to measure end-expiratory lung volume at each step of our protocol, these methods cannot account for tidal recruitment or overdistension (35). Third, whether regional pleural strain is a good surrogate for regional pulmonary strain remains to be determined. Alveolar deflation kinetics in a small animal model suggest that differences between the lung’s periphery and core regions exist and that these may be altered by disease states (36, 37). On the other hand, lung lesions in another VILI animal model developed predominantly in the subpleural regions (38). Further work will need to be done to determine the clinical usefulness of lung US-measured regional pleural strain. Fourth, flow, pressure and volume curves were not recorded simultaneously with each image acquisition. As such, the precise beginning and end of each respiratory cycle wasn’t exactly known which may have affected our results. The exact tidal volume administered while acquiring cineloops may also have been slightly different from the tidal volume recorded beforehand. Fifth, physiological lung sliding and rib shadows preclude measuring the same pleural surface throughout the respiratory cycle. While we computed strain between two consecutive frames of a given cineloop, the pleura imaged in the ROI at end-inspiration was likely different from the pleura in the ROI at end-expiration. As such, strain measures likely represent average strain values from a pulmonary neighborhood. Sixth, the elastography parameters calculated from the Lagrangian speckle model estimator reflect dynamic pleural strain (pleural deformation from regional tidal volume/gas volume at set positive end-expiratory pressure) and not total pleural strain (pleural deformation from total gas volume/gas volume at functional residual capacity). While experimental evidence suggests dynamic strain is the most important strain component leading to the development of VILI (39–41), the clinical usefulness of lung US-measured regional pleural strain will need to be verified in future studies. Seventh, patients included in this proof-of-concept study were highly selected as we excluded patients with any lung pathology and obese patients. While our first sensitivity analysis that excluded suboptimal images did not show different results (Supplementary Table 1 and Supplementary Figure 5), our results’ external validity remains to be demonstrated and future studies will have to include obese patients and patients with various lung pathologies. Finally, like all ultrasonographic exams, lung ultrasonography is operator dependent. Fortunately, because of lung ultrasonography’s shallow learning curve, it is possible to train a clinician rapidly. As few as 10 supervised exams have been found to be necessary when assessing simple pathologies (42) although interpreting more complicated quantitative aspects of lung US requires a 2-month period and 25 supervised exams (43).

In conclusion, measuring regional pleural strain by ultrasound elastography is feasible. A significant dose-response relationship between tidal volume and lateral absolute shear, lateral absolute strain and Von Mises strain was observed, supporting the hypothesis that regional pleural strain can be measured at the bedside with non-invasive lung ultrasound. Further work will be required to compare elastography parameters to gold-standard strain measurements and how they can be used to tailor mechanical ventilation in the operating room and the intensive care unit to improve patient outcomes.

Due to national regulations in the Province of Quebec (Canada), health medical data cannot be made available publicly. However, access to the research dataset is possible for research purpose after appropriate privacy agreements between research parties have been completed. Data access requests may be sent to the corresponding author MG, martin.girard@umontreal.ca, or directly to the CHUM REB (ethique.recherche.chum@ssss.gouv.qc.ca). The R code will be available upon request to the corresponding author.

The studies involving human participants were reviewed and approved by the Comité d’éthique de la recherche du CHUM. The patients/participants provided their written informed consent to participate in this study.

MG, YC, GC, and AD designed the study. MG and SG collected the data. MG, M-HRC, MC, GC, and AD analyzed the data. MG wrote the manuscript. All authors have read and approved the final manuscript.