The median fitting errors in healthy subjects during inflation and deflation were less than 3.1%. Similar to healthy subjects, the model fits well for ARDS subjects with median absolute percentage error during inflation and deflation less than 4.7%. There is a noticeable high median fitting error for ARDS subject 9 at PEEP 5cmH
2
O, at 27.32% during inflation. The model was not able to capture these specific physiological conditions at low PEEP. In particular, this case can be associated with the effect of Auto-PEEP distorting the actual lung condition [9]. The recruitment model fits better when Subject 9 is ventilated at higher PEEP (P < 0.005) compared to lower PEEP. However, the relatively low median error overall subjects indicates the model is capable of capturing fundamental mechanics of both healthy and ARDS lungs.
Tables 2-3 show the estimated mean TOP and TCP for all the healthy and ARDS subjects. In healthy subjects, the overall mean TOP is decreased with increasing PEEP. Mean TCP increases with increasing PEEP. The TOP and TCP distribution shift of a subject during PEEP increase is observed in Figure 6 (Bottom), and are capturing the recruitment as expected.
Similar mean TOP and TCP trends are also observed in ARDS subjects. However, an overall higher TOP is observed compared to healthy subjects, which is also expected for an ARDS lung. The overall higher mean TOP indicates that the ARDS lung consists of relatively more collapsed alveoli and higher pressure is needed to recruit the collapsed alveoli.
Healthy lungs normally consist of only opened or recruited lung units, and a model based on the concept of recruitment may not be applicable. However, in a healthy anesthetized and sedated subject, pulmonary atelectasis can be observed, but it is less severe compared to an ARDS lung and can be easily recruited [12–14]. Thus, during inflation, relatively lower pressure is needed to ventilate the healthy “collapsed” lung compared to ARDS lung. Therefore, for a given tidal volume, the area within the PV curve for a healthy lung should be smaller than ARDS lung. Equally, the healthy lung is less heterogeneous and the lower SD will keep the PV loop area smaller. Figure 7 shows a clear comparison of a healthy and ARDS PV curve, in which the ARDS PV curve has greater area than the healthy PV curve and correspondingly higher SD for this Subject 5 in Table 4. The change thus shows the expected higher work of breathing in the heterogeneous ARDS lung.
Comparing the healthy and ARDS state, mean TOP for healthy lungs are lower when compared to ARDS lungs in Figure 7 (Bottom Left). A healthy lung is a less heterogeneous lung and the effect of superimposed pressure to alveoli is less detrimental. As suggested earlier, a healthy lung is normally open, which results in a lower mean TOP. Thus, the model captures the fact that, for the same subject at a healthy and ARDS state, higher pressure is required to recruit and open the lung. The inter-subject variability in this behavior is evident in Figure 8. Overall, these model results match clinical observation and expectation, which further validates the model.
The deflation curve remains unchanged in ARDS compared to healthy subjects, as shown in Figure 7 (Top), which results in relatively no change in TCP, as seen in Figure 7 (Bottom Right) and Table 3. Hypothetically, mean TCP should be higher in the ARDS state compared to the healthy state [10, 11]. ARDS lung units are more unstable and vulnerable to collapse. Thus, higher pressure is required to retain recruitment. However, this hypothesis was neither observed nor apparent in these results. Only a small increase in TCP is observed during ARDS state compared to healthy state as shown in Figure 7.
The DSG for the ARDS subjects are shown in Figure 8. It is observed that all 3 subjects experienced different SD and TOP increase when transitioning from healthy to ARDS state. In particular, Subject 5 has a relatively small increase in both SD and TOP between healthy and ARDS state. Subject 6 had very large increase in SD (heterogeneity) but less change in TOP (Collapsed lung units). Subject 9 had a very high TOP change (Lung collapse) but minimal changes in SD (Heterogeneity). These results show the diversity in the impact of the ARDS induced.
It is known that ARDS induced in animal model using oleic acid are highly variable [15]. A small variation in ventilation and hemodynamic management during preparation, time and dosage may alter the severity or extensiveness of the lung injury, resulting in different pathophysiological consequences [15–18]. That behavior is clearly evident in these results.
Importantly, this research focuses on minimal model performance in healthy and ARDS lungs. Combining the DSG for all 3 subjects, as shown in Figure 8, the healthy subjects have overall lower TOP and SD than in the ARDS state. This finding suggests that the DSG application is not limited to patient-specific disease state tracking, and it is possible to be expanded into population monitoring. Capturing 3 different ARDS respiratory mechanics or pathophysiological consequences, thus encourages the model’s application in clinical setting, where the presentation of ARDS and its evolution over time and treatment can be variable.
This DSG application is unique and observing DSG shifts should provide useful information for clinical decision support. For example, patients who are grouped in Panel D (High TOP, low SD), have a less heterogeneous lung, but with overall higher lung unit opening pressure. For example, it is hypothesized that a high PEEP can be used in MV to recruit overall collapsed lung units and improve gas exchange [19, 20]. For patients who are grouped in Panel A (Low TOP, high SD), ventilation modes with 2 PEEP levels (Bi-Level PEEP ventilation, airway pressure release ventilation (APRV)) can reduce cyclic opening and collapse of lung units and improve patient outcome [21–23]. Tracking patient DSG with time will also show the effect and patient’s response to specific treatment. In this research, the effect of oleic acid can be seen in increase of TOP and SD. However, the exact limits of these groupings remain to be determined, although it does not affect the ability to track patient condition and response to therapy as in Figure 8.
Overall, the difference of mean TOP and SD between the healthy and ARDS state can be identified using the minimal model. The application of minimal model is not limited to the diseased lung, and allows comparison between healthy and ARDS lungs, and thus encourages its application and future investigation in the ICU to monitor patients-specific condition to guide MV therapy. An overall down shift of mean TOP and/or lowered SD will indicate that the lung recovering for injurious state. In contrast, an up-shift of TOP and/or SD, will show that the lung is more injured. This unique pair of metric thus provides the ability to track the disease state from healthy to injured state and vice-versa. However, mean TCP appears to have little change between healthy and ARDS state, indicating that the TCP parameter was less significant in this clinical use.