After instructing patients in vocal techniques (pitch and loudness), the vowels of /a/,/i/, and/u/were assessed at varying pitch and loudness levels through a visual analog scale (VAS). Task1: To observe different vocal pitches, patients were instructed to relax their whole body and pronounce /a/,/i/, and/u/at low (VAS 2-3) and high pitches (VAS 7-8) with the same loudness (Bane et al., 2023; Awan et al., 2013; Awan et al., 2012). Task 2: To observe different vocal loudness, patients were instructed to pronounce /a/,/i/, and/u/at low (VAS 2-3) and high loudness (VAS 7-8) with the same pitches. Task 3: The same vocalization loudness of 60 dB was selected, and the patient was instructed to continue for more than 5 s to reach the tension-time threshold. The visual feedback interface displays real-time loudness. The patients had rest time between tasks and communicate fully to one another to ensure that patients are relaxed before pronouncing. The Borg-CR10 breathlessness scale was used to assess the degree of dyspnea of the patients after each cycle of /a/,/i/,/u/. Two tasks were carried out at different periods, and the patients were assured of adequate rest before each task. A decibel meter measured at least 15 dB between low and high levels for accurate loudness difference. To assess different pitches, the patients initially produced a comfortable pitch while maintaining a comfortable loudness and then shifted to a higher base pitch (at least +5 semitones) while monitoring the pitch by using online tuning software (Wang and Yiu, 2023). Throughout the process, the voice was required to maintain a stable loudness and pitches and measured approximately 20 cm away from the participant. Data acquisition concluded when the voice reached a weak level (decrease of more than 5 dB or 2 semitones).

The Delsys Trigno™ wireless system (Delsys, Natick, MA) and four attached electrodes captured EMG signals at a sampling rate of 2,000 Hz. Wireless electrodes were placed at specific anatomical locations to ensure accurate measurements: the middle and lower 1/3 of the SCM, the junction of the PARA, the 7thIC, and 3 cm above the umbilicus in the position of the RA (Figure 1). Before electrode placement, the skin was meticulously cleaned with a medical alcohol pad to ensure optimal signal acquisition (Liu et al., 2019; da Fonsêca et al., 2019; Cabral et al., 2018). All surface EMG electrodes were positioned on the right side of the body for consistency (Ramsook et al., 2016). After data acquisition, all sEMG signals were analyzed using EMG works analysis software (Delsys, Natick). The sEMG signals recorded were filtered by a 20–450 Hz band-pass Butterworth filter. The signals were then segmented using a root-mean-square (RMS) value calculated based on a 100 ms moving window (Stepp, 2012) (Figure 1). The sEMG signal was calibrated as a percentage of the sEMG signal at maximum voluntary contraction (MVC) (Nguyen et al., 2020). NRD was used to represent RMS%MVC. During the performance of the MVC task, the subjects were instructed to exert their best effort by conducting three maximum breath tests (Cabral et al., 2018).

Statistical analysis used SPSS software v26.0 and GraphPad Prism v9.5.1. Data with normal distribution were presented as mean ± standard deviation (SD), while data with a skewed distribution were expressed as median ± interquartile range (IQR). Repeated-measures analysis of variance (RM-ANOVA) was utilized to analyze the EMG data of respiratory muscles and Borg scale across different tasks. Correlation analysis was conducted using Pearson correlation. A 95% confidence interval was established, and the significance level was set at 0.05.

We found no difference at the NRD of SCM (F (2, 120) = 0.116, P = 0.890), PARA (F (2, 120) = 0.034, P = 0.967), 7thIC (F (2, 120) = 0.755, P = 0.473), and RA (F (2, 120) = 0.019, P = 0.982) during the high pitches task compared with the low pitch task (Table 2).

The study showed no significantly higher NRD in patients with high loudness compared to those with low loudness in SCM (F (2, 120) = 1.300, P = 0.276) (Figure 2A). Furthermore, COPD patients with high loudness exhibited significantly higher NRD in the PARA (F (2, 120) = 55.27, P < 0.001) (Figure 2B), 7thIC (F (2, 120) = 59.08, P < 0.001) (Figure 2C), and RA (F (2, 120) = 39.56, P < 0.001) (Figure 2D) when compared to those with low loudness.

No differences in NRD were found for SCM (F (2, 120) = 0.309, P = 0.735), PARA (F (2, 120) = 0.058, P = 0.944), 7thIC (F (2, 120) = 0.051, P = 0.944), and RA (F (2, 120) = 2.128, P = 0.124) across vowel tasks (Table 2).

Borg dyspnea scores significantly differed across tasks [F (3, 160) = 66.47, P< 0.001]. We found no statistically significant difference in Borg dyspnea scores under low pitch and low loudness. The difference between Borg dyspnea scores under the high sound task was not statistically significant. However, Borg dyspnea scores were significantly higher after the high-loudness task than after the low-loudness task (P< 0.001) and after the low-pitched task (P< 0.001). Borg dyspnea scores were also significantly higher after the high-pitched task than after the low-loudness task (P< 0.001) and after the low-pitched task (P< 0.001) (Figure 3).

Correlation analysis of expiratory muscle RMS measured after harmonization of pitch with baseline patient data showed that loudness was unrelated to baseline patient status. However, the RMS of the loudness control task was significantly negatively correlated with MEP (r = −0.671, P < 0.001) and MIP (r = −0.571, P < 0.001). Higher respiratory muscle strength was associated with lower RMS (Figure 4).

Categorization of vowels into high vowels and low vowels, along with mid vowels, is a well-established concept in linguistics. Subglottic pressure is produced by respiratory and laryngeal muscles, which is necessary for voice change (Traser et al., 2020). However, no significant difference was found in the NRD of respiratory muscle under different pitches and vowels. Plexico and Sandage (2012) showed that under any evaluation frequency, the threshold pressure value was not significantly different among the three consonant-vowel sequences, similar to the present results. However, Pettersen et al. found significant differences in the activation of intercostal, lateral abdominal, and rectus abdominus with different pitch and loudness levels. The vocalization task in Pettersen’s study did not control for confounding factors, such as sound loudness. By contrast, loudness was controlled in this study, which may be the reason for the difference in the results (Pettersen, 2005). In addition, high-pitched voices produced prokinetic symptoms despite no significant respiratory muscle activation. Correspondingly, Wang and Yiu (2023). Showed that the activation level of suprhyal muscle was different with different vowels due to the different shapes of the mouth and the position of the tongue. Recruitment of laryngeal muscles increased significantly with increased pitch (Zhu et al., 2022). This finding suggests that perilaryngeal muscles rather than respiratory muscles produced pitch changes. This phenomenon is consistent with the view that the original power of the respiratory muscle produces sound and that the larynx and mouth are mainly used to modify the sound (Körner and Strack, 2023; Herbst, 2017). In conclusion, this study confirms that different pitches and vowels in different training programs do not directly affect respiratory muscle training effects.

The vocal effort produced significantly greater subglottic pressure during maximum-effort speech (Rosenthal et al., 2014), which may have increased the load during the expiratory phase. However, not all expiratory muscles were significantly associated with loudness. In the loudness control task, the lower ribcage embodied vocalization preferentially because only the 7thIC showed differences while PARA and RA did not. The lack of coordination of abdominal vocalization during the vocalization state in patients with chronic obstructive pulmonary disease was unexpected. We observed a more generalized chest breathing habit in patients with COPD, which may explain the lack of pronounced abdominal muscle vocalization. Some research suggests that SCM muscles may help stabilize human vocalizations (Van Houtte et al., 2013). Pettersen et al. (2005) showed that when healthy ordinary people (professional or non-professional singers) participated in vocalization training and performed vocalization content with different loudness and pitch levels, the activities of the sternocleidomastoid muscle and trapezius muscle increased; this effect was more noticeable when the respiratory demand was strong. Similarly, our observational research showed that the NRD of respiratory muscle was not significantly higher than that of low-loudness vocalizations in SCM in patients with COPD.

Previous studies have shown that comprehensive pulmonary rehabilitation, based on standard protocols, can improve dyspnea and quality of life in patients (Troosters et al., 2023). However, there are still inconsistencies in vocalization training programs for COPD patients (Fang et al., 2022). This study offers clinical implications for the implementation of vocalization training. Regarding the physiological effects on patients, Fu et al. showed that collective singing and vocal training improved the maximum expiratory pressure and exercise ability of older people in the community (Fu et al., 2018). Lord et al. (2012) reported that the objective physiological state of patients with COPD did not change after vocalization training. However, the content and intensity of singing are not disclosed in vocalization studies, which may be the reason for the differences in the outcome indicators (Patel et al., 2022; Selickman and Marini, 2022). In addition, correlation analysis showed that vocalization of the same loudness produced great stimulation in patients with low respiratory muscle strength. This finding is consistent with the theory that the intercostal and abdominal muscles provide vocal expiratory support (Pettersen et al., 2005; De Troyer and Boriek, 2011). Muscle control and improvement are required to ensure an intensity threshold for training; according to the results of this study, vocalization at a lower intensity may not be sufficient to engage the respiratory muscles fully. At the same time, the degree of dyspnea after the high-loudness task was higher than that after the high-pitched task, while the degree of dyspnea between the low-loudness task and the low-pitched task was not significantly different. Hence, home oxygen therapy, breathing exercises, and other treatments that can improve dyspnea can be combined with vocalization training for improved clinical outcomes. Personalized vocalization prescriptions for patients with COPD should consider the respiratory muscle state of patients rather than simply pursuing vocalization loudness.

Although the study examined the NRD of respiratory muscles to different vocalization tasks, this study still has some limitations. First, this study only addressed the physiological effects of vocal tasks and did not address the psychological effects of vocal training as an artistic engagement. Second, all syllable pitches and loudness are difficult to study because of the patients’ limited tolerance. Relatively simple /a/, /i/, and/u/were chosen for this study, considering vocal teachers’ suggestions and previous studies. Finally, participants included patients with COPD only who had performed primary vocal exercises to study the effect of training on uninitiated patients. Therefore, the results of the article cannot be generalized to patients who have received full vocalization training.