Visualizing pursed lips breathing of patients with chronic obstructive pulmonary disease through evaluation of global and regional ventilation using electrical impedance tomography

This study was approved by the ethics committee of the Fourth Military Medical University (KY20224101-1) and was conducted in accordance with the principles embodied in the Declaration of Helsinki and in accordance with local statutory requirements. All patients were asked to provide signed informed consent forms prior to the study. In-patients with acute exacerbation COPD (AECOPD) from 12/2022 to 8/2023 were screened prospectively. The inclusion criteria were: (1) younger than 90 years old; (2) free from other significant pulmonary disease with obvious symptoms; (3) difficult to complete deep breathing due to cough and other reasons; (4) AECOPD was under control and the patients were ready to be discharged; (5) scheduled for PLB training and novice to PLB. Exclusion criteria included: difficult to communicate with; chest skin injury against fixation of EIT electrode belt and other contraindications to use pulmonary EIT (e.g. cardiac pacemaker).

According to the order of inclusion, the patients were sequentially divided into two groups in an alternate order: conventional group and EIT guided group. Before the start of the formal experiment, the patient was instructed to maintain a sitting position and PLB was taught to each patent by an experienced physiotherapist. The breathing procedures for PLB maneuver were defined as follows (Garrod et al 2005). First, the respiratory accessory muscles were kept relaxed, such as neck and shoulder muscles. Second, air was inhaled deeply and slowly through the nose. Third, air was exhaled slowly at least four times longer than inspiration by pursing or puckering lips, meanwhile cheek not puffed out.

In the conventional group, PLB was taught to the patient by the following steps: First, the role of PLB was explained in improving respiratory symptoms to make the patient understand the significance of PLB and thus to enhance their learning motivation. Second, a lighted candle was placed about 25 cm in front of the patient, which was used to assess the expiratory flow, i.e. during the exhalation, the flame continued to move slightly, but it could not be blown out, as shown in figure 1(a). Third, the points of PLB maneuver were repeatedly explained to the patent with multiple maneuver demonstration by the physiotherapist in order to improve his/her PLB performance. The entire teaching process was limited for 10 min.

Zoom In Zoom Out Reset image size

In the EIT guided group, the significance of PLB was first explained, as in the conventional group. Second, the global and regional EIT parameters were explained to the patient. Third, the points of PLB maneuver were also stated to the patent with multiple maneuver demonstration by the physiotherapist, and simultaneously the patient improve his/her PLB maneuver by real-time observing the EIT parameters, as shown in figure 1(b). The entire teaching process was also limited for 10 min.

After learning, each subject was familiar with PLB and was asked to complete a final test, involving the following actions in sequence: taking quite breathing (QB) for 1 min, breathing as deep as possible for five cycles, returning QB for 1 min, performing 15 respiratory cycles of PLB without the feedback from EIT, and restoring QB for 1 min. During the whole test, pulmonary EIT data were continuously measured and saved.

Finally, another experienced physiotherapist, who did not know which group the patient belonged to, rated the PLB performance based on his subjective evaluation according to the items used in the clinical practice, as listed in table 2 (i.e. without the help of EIT). Then he rated the PLB performance again after carefully viewing the EIT results (i.e. with the help of EIT).

At the end of the study before they were discharged, all patients were training to manage PLB maneuver regardless of their grouping.

The skin of the chest in the 4th–5th intercostal region was cleaned by using medical alcohol to reduce the effect of the corneum on the electrode-skin conductivity. Next, considering the larger magnitude of respiration during implementing PLB, a specially designed electrode belt with the appropriate length was selected by measuring the chest circumference of the patient to enhance the stability of the electrode-skin interface, which equally spaced 16 electrodes involving 12 fixed conductive rubber electrodes and 4 buttons located on both sides of the spine and the sternum for connecting the disposable ECG electrodes, as shown figure 2(a). The 12 rubber electrodes and 4 ECG electrodes are treated the same in the system and used for current injection and voltage measurements consecutively. Finally, the electrode belt was attached to the chest in one transverse at the level of the 4th–5th intercostal space at the parasternal line and was connected to the EIT system through an electrode wire, as shown figure 2(b).

Zoom In Zoom Out Reset image size

Raw EIT data was collected with a commercial EIT system (VenTom-100, MidasMED Biomedical technology, Suzhou, China) at an acquisition rate of 20 frames s⁻¹ with working frequency of 50 kHz and signal-to-noise >70 dB. In order to improve the sensitivity in the central region of the chest, the data acquisition mode of opposite excitation-adjacent measurement was chosen to inject the current of 1 mArms through pairs of electrodes opposite each other and measure the voltage on other adjacent electrodes (Yang et al 2021a). Totally 16 × 12 voltage values per frame are used to reconstruct one image. Details of EIT image reconstruction are summarized in the online supplement.

2.4.1. Assessment of global ventilation during performing PLB

The sum of all pixel values of each EIT image was determined as the global impedance at a certain moment to evaluate the global ventilation at one time point because of the good correlation between impedance variation and ventilation volume change. Thus, the global impedance changes ${\rho }_{t}$ calculated from time series EIT images was used to reflect the variation of global ventilation against time, that is, ${\rho }_{t}=\displaystyle {\sum }_{i=1}^{1024}{\rm{\Delta }}{{\boldsymbol{\rho }}}_{t}^{i},$ where ${\rm{\Delta }}{{\boldsymbol{\rho }}}_{t}^{i}$ denotes the value of $i{\rm{th}}$ pixel at the time point $t.$ Based on the special requirements of PLB maneuver, the following metrics were proposed to assess the characteristics of PLB maneuver.

2.4.1.1. Inspiratory depth

This metric assesses the degree of the inspiratory force, which was calculated by the percentage of inspiratory volume during performing PLB to that during performing deep breathing, that is, ${\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}}}/{\rho }_{{\rm{V}}{\rm{T}}}^{\max },$ where ${\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}}}$ and ${\rho }_{{\rm{V}}{\rm{T}}}^{\max }$ denote the tidal impedance changes during PLB and deep breathing, respectively, as shown in figure 3. ${\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}}}$ and ${\rho }_{{\rm{V}}{\rm{T}}}^{\max }$ were obtained by calculating the average of tidal impedance changes of all breathing cycles during performing PLB and deep breathing, respectively, i.e. ${\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}}}={\rm{m}}{\rm{e}}\mathrm{an}\left({\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}},1},{\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}},2},\ldots ,{\rho }_{{\rm{V}}{\rm{T}}}^{{\rm{P}}{\rm{L}}{\rm{B}},n},\ldots \right)$ and ${\rho }_{{\rm{V}}{\rm{T}}}^{\max }={\rm{m}}{\rm{e}}\mathrm{an}\left({\rho }_{{\rm{V}}{\rm{T}}}^{\max \,,1},{\rho }_{{\rm{V}}{\rm{T}}}^{\max \,,2},\ldots ,{\rho }_{{\rm{V}}{\rm{T}}}^{\max \,,n},\ldots \right),$ where ${\rho }_{\mathrm{VT}}^{\mathrm{PLB},n}$ was the tidal impedance change of $n{\rm{th}}$ respiratory cycle during PLB and ${\rho }_{{\rm{V}}{\rm{T}}}^{\max \,,n}$ was the tidal impedance change during deep breathing.

Zoom In Zoom Out Reset image size

2.4.1.2. Ratio of expiratory time to inspiratory time

This metric assesses the time allocation of the whole respiratory cycle during PLB, which was calculated by the percentage of expiratory time to inspiratory time, that is, ${t}_{\exp }/{t}_{{\rm{i}}{\rm{n}}{\rm{s}}}.$ The averages of the percentage of expiratory time to inspiratory time during all breathing cycles were taken as the final value, i.e. ${\rm{m}}{\rm{e}}{\rm{a}}{\rm{n}}\left(\tfrac{{t}_{\exp }^{1}}{{t}_{{\rm{i}}{\rm{n}}{\rm{s}}}^{1}},\tfrac{{t}_{\exp }^{2}}{{t}_{{\rm{i}}{\rm{n}}{\rm{s}}}^{2}},\ldots ,\tfrac{{t}_{\exp }^{n}}{{t}_{{\rm{i}}{\rm{n}}{\rm{s}}}^{n}},\ldots \right).$

2.4.1.3. Expiratory flow

This metric measures the speed of exhalation during PLB, which was calculated by the ratio of inspiratory volume to expiratory time. To eliminate the individual differences, the relative inspiratory depth was used to reflect the inspiratory. Thus, normalized expiratory flow could be obtained by $\mathrm{mean}\left({F}_{1},{F}_{2},\ldots {F}_{n},\ldots \right),$ where ${F}_{n}=\tfrac{{\rho }_{\mathrm{VT}}^{\mathrm{PLB},n}/{\rho }_{\mathrm{VT}}^{\mathrm{PLB}}}{{t}_{\exp }^{n}},$ and ${\rho }_{\mathrm{VT}}^{\mathrm{PLB},{\rm{n}}}$ was the tidal impedance change of $n{\rm{th}}$ respiratory cycle.

2.4.1.4. Expiratory uniformity

This metric depicts the uniformity of exhalation during PLB, which was calculated by the variation of exhaled volume within unit time throughout the expiratory phase, that is $\mathrm{mean}\left(\ldots ,\tfrac{\mathrm{std}\left(\ldots ,{\rho }_{{\rm{\Delta }}t,m}^{\mathrm{PLB},n},\ldots \right)}{\mathrm{mean}\left(\ldots ,{\rho }_{{\rm{\Delta }}t,m}^{\mathrm{PLB},n},\ldots \right)},\ldots \right),$ where denoted the global impedance change within $m{\rm{th}}$ unit time during $n{\rm{th}}$ respiratory cycle, as shown in figure 3. In this study, the unit time was defined as 200 ms.

2.4.1.5. Respiratory stability

This metric describes the degree of respiratory condition during PLB, which as calculated by the variation of global impedance at the end of exhalation, that is, $\tfrac{\mathrm{std}\left(\ldots ,{\rho }_{\mathrm{end}-\exp }^{\mathrm{PLB},n},\ldots \right)}{\mathrm{mean}\left(\ldots ,{\rho }_{\mathrm{end}-\exp }^{\mathrm{PLB},n},\ldots \right)},$ where represents the global impedance at the end of exhalation during $n{\rm{th}}$ respiratory cycle, as shown in figure 3.

Based on the global ventilation, the physiotherapist rated the PLB performance according to the items used in the clinical practice, as listed in table 1.

Table 1. Rating score table for assessment of pursed lips breathing maneuver.

2.4.2. Assessment of regional ventilation during performing PLB

In order to evaluate the regional ventilation during PLB implementation, the functional EIT (fEIT) images were constructed by employing time series EIT images to quantify the specific local ventilatory features. The regression method for global-regional impedance change was adopted in this study because of its robustness to signals with different phase information. For each subject, the fEIT image of PLB (${\bf{Z}}$) was computed, which consists of regression coefficients of the following linear regression formular,

$$\begin{eqnarray*}&&{\rm{\Delta }}{\rho }^{i}\left(t\right)={\alpha }_{i}\cdot \displaystyle \sum _{i=1}^{1024}{\rm{\Delta }}{{\boldsymbol{\rho }}}_{t}^{i}+{\beta }_{i}+{\chi }_{i}={\alpha }_{i}\cdot {\rho }_{t}+{\beta }_{i}+{\chi }_{i},\end{eqnarray*}$$

where ${\rm{\Delta }}{\rho }^{i}\left(t\right)$ denotes the time-dependent impedance change of $i{\rm{th}}$ pixel of the raw time series EIT images; $\alpha$ and $\beta$ are the regression coefficients; $\chi$ is the fitting error.

Because PLB is a breathing pattern subjectively controlled by the patient, the regional ventilation distribution may be affected by the respiratory muscles. To measure the effects, ventral to dorsal side ratio ($V\mathrm{to}D$), right to left lung ratio ($R\mathrm{to}L$), as well as center of ventilation (CoV) were calculated, respectively. The $VtoD$ was obtained by computing the ratio of sum of pixels in the dorsal side to that in the ventral side, i.e. $\displaystyle \sum _{i{\in }\mathrm{ventral}}{{\bf{Z}}}^{i}/\displaystyle \sum _{i{\in }\mathrm{dorsal}}{{\bf{Z}}}^{i}.$ The $R\mathrm{to}L$ was obtained by $\displaystyle \sum _{i{\in }\mathrm{right}}{{\bf{Z}}}^{i}/\displaystyle \sum _{i{\in }\mathrm{left}}{{\bf{Z}}}^{i}.$ CoV, another parameter depicting ventilation distribution, was calculated by the impedance value weighted with a position in the anteroposterior coordinate (Frerichs et al 1998).

Moreover, considering the prolonged exhalation during PLB, the regional expiratory time constants were calculated to evaluate the regional airflow. The time constant of each pixel was calculated by fitting an exponential curve to the expiratory impedance signal:

$$\begin{eqnarray*}&&{\rm{\Delta }}{{\boldsymbol{\rho }}}^{i}={\rm{\Delta }}{{\boldsymbol{\rho }}}_{0}^{i}\cdot {e}^{\displaystyle \frac{-t}{{\tau }^{i}}}+C,\end{eqnarray*}$$

where ${\tau }^{i}$ is the time constant of $i{\rm{th}}$ pixel and $C$ is the end-expiratory impedance. In this study, the time constants of all pixels within the lung area were calculated to obtain the EIT-derived time constant image, in which the lung area consisted of the pixels with the value >20%*maximum of the fEIT image. For each subject, the final EIT-derived time constant EIT image was obtained by computing the average of the time constant EIT images of all breath cycles during PLB. Furthermore, the median of the distribution of final EIT-derived time constant EIT image was taken as the final expiratory time constant for each patient.

Statistical analysis was conducted with SPSS 26 (IBM Software, Armonk, NY) and a P value of <0.05 was deemed statistically significant. For global ventilation, the five metrics (inspiratory depth, ratio of expiratory time to inspiratory time, expiratory flow, expiratory uniformity, and respiratory stability) during performing final PBL test and calm breathing in the conventional group were compared with those in the EIT guided group by using one-way ANOVA with post-hoc test. For regional ventilation distribution, the three indexes ($V\mathrm{to}D,$ $R\mathrm{to}L,$ CoV and expiratory time constant) of two groups during PLB and QB were also compared. For rating scores, the differences of rating scores in the two group were first compared, and then the rating scores with and without EIT were also compared by employing Bland–Altman plots and Student t-test.

Figure 4 shows the global ventilation change of one patient in the conventional group and one patient in the EIT guided group during performing QB and PLB. For both groups, the tidal volume during PLB was larger than that during QB whereas the breathing rate was otherwise. Compared to QB, the expiratory time was obviously longer than the inspiratory time during PLB.

Zoom In Zoom Out Reset image size

Figure 5 shows the comparisons of global ventilation parameters between the conventional group and EIT guided group during performing QB and PLB. In terms of QB, none of the global ventilation parameters in the conventional group was significantly different from that in the EIT guided group.

Zoom In Zoom Out Reset image size

For both groups, the inspiratory depth during PLB implementation was significantly larger than that during QB implementation (figure 5(a), P < 0.001). The inspiratory depth during performing PLB in the EIT guided group (0.74 ± 0.09) was greater than that in the conventional group (0.60 ± 0.19, P < 0.001). But in the EIT guided group, all patients had a relatively high level of inspiratory depth whereas there were still 3 patients with an inspiratory depth of less than 0.4. These results were similar to those of the ratio of expiratory to inspiratory time (figure 5(b)). Only 3 patients' ratio of expiratory to inspiratory time during PLB in the EIT guided group was smaller than 4 whereas there were 10 patients in the conventional group.

The expiratory flow during PLB in both groups was significantly smaller than that during QB (figure 5(c), P < 0.001). Expiratory uniformity during PLB was significantly larger than that during QB in the conventional group (figure 5(d), P < 0.05) whereas no significant difference was found in the EIT guided group (P = 0.789).

The respiratory stability during PLB in the conventional group was significantly larger than that during QB (figure 5(e), P < 0.01). In contrast, the respiratory stability during PLB in the EIT guided group was not significantly different from that during QB (P = 0.0541) and it was also significantly lower than that during PLB in the conventional group (P < 0.001).

Figure 6 shows the comparisons of regional ventilation parameters between the conventional group and EIT guided group during performing QB and PLB. During QB, none of the four regional ventilation parameters in both groups was significantly different.

Zoom In Zoom Out Reset image size

During PLB, the ventral to dorsal side ratio increased in both groups (P = 0.125, figure 6(c)), but only a significant difference was observed in the EIT guided group (P < 0.01). No significant difference in right to left lung ratio in both groups was found (figure 6(d)). Compared to QB, CoV significantly decreased during PLB in both groups (figure 6(e), conventional group P < 0.05, EIT guided group P < 0.01). For both groups, the expiratory time constant during PLB was significantly larger than that during QB implementation (P < 0.001) and the expiratory time constant during PLB in the EIT guided group was greater than that in the conventional group (P < 0.001).

Figure 7 shows the analysis of rating scores of PLB performance with and without the help of EIT. Bland–Altman plots analysis suggested that 87.5% and 100% differences were in the 95% limits of agreement for the conventional group and EIT guided group, respectively (figures 7(a) and (b)). The score rated after EIT observation significantly lower than that rate subjectively in both groups (P < 0.05). The score drop in the conventional group (from −5 to −1, −2.68 ± 1.1) was obviously larger than that in the EIT guided group (from −2 to 0, −1.19 ± 0.72).

Zoom In Zoom Out Reset image size

In terms of global impedance, the inspiratory depth and the ratio of expiratory to inspiratory time during PLB were larger than those during QB while the expiratory flow was smaller, which indicated that EIT correctly captured the volume and time characteristics of PLB and QB. These results were consistent with previous studies on pulmonary physiology of PLB. Moreover, significant differences in inspiratory depth and the ratio of expiratory to inspiratory time between the two groups were observed (figure 5(a) and (b)), suggesting that with the help of EIT the patients have better understanding of PLB implementation. No significant difference in expiratory flow between two groups was found (figure 5(c)). We suspected that the patients in the conventional group not only carried out a shorter exhalation (indicated by the ratio of expiratory to inspiratory time), but also performed a shallower breathing (indicated by the inspiratory depth). Thus, expiratory flow (calculated by the ratio of the inspiratory volume to expiratory time) maintained a comparable level comparable to the EIT guided group. The use of candle in the conventional group might also contribute to the steady continuous exhalation. In addition, the expiratory uniformity in the EIT guided group was significantly smaller than that in the conventional group (figure 5(d), P < 0.01), suggesting inhaled gas was exhaled more uniformly through by subjectively controlling the respiratory muscles. This could be obviously observed from the global impedance change, in which the impedance curve during expiration phase in the EIT guided group dropped almost with a fixed slope (figure 4(b)) whereas not in the conventional group (figure 4(a)). Moreover, the respiratory stability during PLB in the EIT guided group was significantly smaller than that in the EIT conventional group (P < 0.01), indicating that with the help of EIT, a more stable breathing depth was maintained by patients in the EIT guided group. In brief, these results demonstrate that compared to the empirical method to teach PLB maneuver, EIT can provide real-time responses to patients during PLB training session from the perspective of the effect of PLB on global ventilation, thereby effectively guiding the patient to master PLB skills.

In terms of regional impedance distribution, compared to QB, the ventral to dorsal side ratio significantly increased and CoV also significantly decreased during PLB implementation. These results may be related to the breathing pattern of PLB. In the experiment, we found that patients performed PLB in a thoracic breathing manner characterized by the apparent bulging of the chest. In the process of thoracic breathing, when the central tendon keeps stable and the diaphragm contracts, the ribs will raise and the chest expands forward and sideways, thus more air entered into the ventral side. However, PLB did not cause a significant difference in the right to left lung ratio, suggesting that the respiratory muscles on both sides had a balanced contraction and relaxation. Moreover, the expiratory time constant of regional ventilation during PLB was larger than that during QB, which might indirectly indicate that PLB avoided the rapid collapse of regional small airway. Therefore, EIT could reveal regional respiratory muscle contractions from the perspective of the effect of PLB on regional ventilation.

A high concordance between subjective rating and rating with the help of EIT was found in both the conventional group and EIT guided group (figure 7), demonstrating the reliability of rating score with EIT. But in both groups, the subjective scores were lower than those rated after observing the EIT parameters. These results may because subjective observation could only roughly evaluate the general situation of PLB maneuver through some body representation, such as estimating inspiratory depth by using the relief of the chest, but not allow accurate judgments of the details of PLB maneuver, such as accurate inspiratory depth (ratio of inspiratory volume to vital capacity), even for an experienced physiotherapist. But EIT could quantify technical details of PLB by providing the proposed metrics of global and regional ventilation during PLB implementation. Moreover, the score drops before and after reviewing EIT data in the conventional group was significantly lower than that in the EIT guided group, further exhibiting that the objective effectiveness of the proposed metrics of global and regional ventilation in assessing PLB. It would be interesting to follow-up whether the differences in the retention of the training between those taught conventionally and those taught with EIT-guidance after a few months. Better design of a study with training interval, follow-up frequency and time points, and possible endpoints should be developed.

PLB is a specific breathing technique, by which preparatory pressure is generated in the airway, making the isobaric point in the airway move to the direction of the alveoli (Vatwani 2019). The small airway is kept dilated when exhaling, which is conducive to the full discharge of residual gas in the lungs and increased tidal volume (Visser et al 2011). The respiratory rate is thus slowed down to facilitate the full diffusion of gas in the alveoli, thereby improving hypoxia and carbon dioxide retention, as well as relaxing the patient's body (Sakhaei et al 2018). In addition, it has been found that PLB training has certain effects on the alleviation of diaphragm fatigue (Breslin 1992). As a result, the patient's breathing difficulties can be eased so that the exercise tolerance is improved (Cabral et al 2015). Although some patients with COPD often carried out PLB spontaneously, teaching patients PLB has been shown to enhance exercise capacity and pulmonary function (Bhatt et al 2013, Borge et al 2014). For example, Bhatt et al assessed the acute effects of volitional PLB on exercise capacity by performing a randomized crossover study comparing 6 min walk test (6MWT) at baseline without PLB with 6WMT using volitional PLB (Bhatt et al 2013). They found that the patients with PLB showed a significant increasement in 6 MW distance compared with spontaneous breathing. Additionally, one high-quality systematic review from Holland et al found a significant positive effect of breathing control exercise based on pooled data analysis with two single RCTs in regard to PLB on breathlessness (Holland et al 2012). Furthermore, in order to encourage the patient with COPD to perform PLB training, music was introduced into PLB teaching (Bonilha et al 2009, Alexander and Wagner 2012, Mcgrath et al 2022). Mackenzie et al proposed a novel PLB training program through music, named MELodica Orchestra for DYspnea (MELODY), to teach PLB patients with COPD, which could be deployed in routine clinical settings (Mcgrath et al 2022). These previous studies demonstrated that teaching PLB to patients could achieve significant positive changes in respiratory and cardiac parameters in COPD patients. In order to make the patient correctly master PLB maneuver as soon as possible, we explored the feasibility of using EIT to assist physiotherapist to teaching PLB to patients with COPD and found that EIT could dynamically capture the details of PLB maneuver so as to improve the efficiency of teaching PLB.

Compared with the traditional method to teach PLB (language communication and subjective judgment), EIT could give the patient and physiotherapist real-time feedback about breathing status, including the global ventilation change (chest global impedance) and regional ventilation distribution (time-series EIT image). On one hand, the responsive information allows the patient to intuitively observe his/her own breathing in a visual way, to quickly understand how the respiratory muscles should be actively controlled to meet the requirements of PLB maneuver. On the other hand, the responsive information permit physiotherapist to visualize the PLB performance of the patient, to rapidly understand the patient's problems in controlling breathing and thus give the corresponding guidance. In brief, EIT create a visual bridge of communication between the patient and the physiotherapist in teaching PLB maneuver to the patient.

According to the characteristics of PLB maneuver, five parameters from global ventilation were proposed to quantitatively assess the details of PLB maneuver, including inspiratory depth corresponding to 'deep inhalation', ratio of expiratory to inspiratory time and expiratory flow corresponding to 'slow exhalation', expiratory uniformity corresponding to 'uniform exhalation', as well as respiratory stability corresponding to 'stable end-expiratory lung volume'. Additionally, four parameters from regional ventilation were proposed to quantitatively evaluate the respiratory muscles during PLB implementation, including ventral to dorsal side ratio and central of ventilation corresponding to 'force way of respiratory muscles', right to left lung ratio corresponding to 'force balance of respiratory muscles', as well as expiratory time constant corresponding to 'force duration of respiratory muscles'. Our results showed that there were significant differences in these indexes between QB and PLB or between the conventional group and EIT guided group. The results demonstrated the scientific rationality and objective practicability of these metrics.

The elastic EIT electrode belt around the chest may affect the subject's lung function. In a previous study, the effect was investigated (Zhang et al 2020). The influence is not neglectable for patients with respiratory muscle weakness but is limited for COPD.

In this study, we established an EIT image reconstruction model based on a real CT image of one male to calculate the reconstruction matrix for reconstructing the time-series EIT images of all patients. In theory, individualized reconstruction model should be used to perform image reconstruction by employing the individualized CT image for each patient. But the absence of CT scans of some patients resulted in an inability to perform personalized EIT image reconstruction. Zhao et al evaluated the effect of reconstruction model on global ventilation and regional ventilation from EIT images by employing the circular model and individualized model (Zhao et al 2014). They concluded that although there were some deviations in the shape and position of the ventilated regions in the EIT images, the EIT image the parameters developed for EIT images generated with individualized model were comparable with those with circular models. Thus, the global and regional ventilation calculated by using the unified reconstruction model was reliable in this study.

Additionally, a software incorporating the analysis of the global and regional ventilation parameters proposed in this study should be developed to further validate the practicability of EIT in teaching PLB to the patient with COPD in routine clinical practice.

The severity of COPD in the conventional group appeared to be slightly worse than in the EIT-guided group based on the FVC parameters, although the difference was not statistically significant. This lack of statistical significance may be attributed to the limited number of participants in each group. Nevertheless, the majority of our evaluations were conducted through intra-patient comparisons, which minimized the impact of any disparities in COPD severity between the two groups prior to treatment.

Using EIT as a form of biofeedback was quite novel and due to the lack of relevant information, we did not calculate an a-priori sample size to detect certain effects. Instead, we screened all the in-patients with AECOPD within the study period and included subjects who were eligible based on the inclusion and exclusion criteria. The observed effects in the current study shall be used to design further randomized-controlled study to validate the efficacy of EIT-guided training.