Impact of Subharmonic and Aperiodic Laryngeal Dynamics on the Phonatory Process Analyzed in Ex Vivo Rabbit Models

Normal voice is characterized by periodic oscillations of the vocal folds. On the other hand, disordered voice dynamics (e.g., subharmonic and aperiodic oscillations) are often associated with voice pathologies and dysphonia. Unfortunately, not all investigations may be conducted on human subjects; hence animal laryngeal studies have been performed for many years to better understand human phonation. The rabbit larynx has been shown to be a potential model of the human larynx. Despite this fact, only a few studies regarding the phonatory parameters of rabbit larynges have been performed. Further, to the best of our knowledge, no ex vivo study has systematically investigated phonatory parameters from high-speed, audio and subglottal pressure data with irregular oscillations. To remedy this, the present study analyzes experiments with sustained phonation in 11 ex vivo rabbit larynges for 51 conditions of disordered vocal fold dynamics. (1) The results of this study support previous findings on non-disordered data, that the stronger the glottal closure insufficiency is during phonation, the worse the phonatory characteristics are; (2) aperiodic oscillations showed worse phonatory results than subharmonic oscillations; (3) in the presence of both types of irregular vibrations, the voice quality (i.e., cepstral peak prominence) of the audio and subglottal signal greatly deteriorated compared to normal/periodic vibrations. In summary, our results suggest that the presence of both types of irregular vibration have a major impact on voice quality and should be considered along with glottal closure measures in medical diagnosis and treatment.

pathologies [26]. These three types of vibrations have been systematically investigated in various studies over the years [27][28][29]. However, not all data can be collected easily in everyday clinical practice from humans. In particular, subglottal measurements cannot (or only under severe conditions) be obtained in vivo; ex vivo experiments on larynges are necessary [30].
Voice production in humans and mammals is very similar [31]; therefore, animal laryngeal studies have been carried out from which conclusions are drawn regarding human physiology. In the literature, voice production of animals and the comparison with humans has already been examined many times. Sheep, canines [32,33], porcines [34,35], and rabbits [36] are particularly suitable for comparison with humans. The first three are used, especially because of the similar laryngeal dimensions to humans, whereas the comparison with rabbit larynges is suitable, because the tissue properties are similar to those of the human larynx [36,37]. Despite this fact, only a few studies of rabbit phonation have been performed. Maytag et al. (2013) [38] suggested a method for reliable extraction of phonatory, acoustic and videokymographic data of ex vivo rabbit larynges, as this form of data collection from ex vivo rabbit larynges was not yet sufficiently investigated. They found that the rabbit data was similar in intralaryngeal variability to canine laryngeal data. In a study published in 2017, Mills et al. [7] made adaptations to an excised booth, primarily used for canine larynges, to examine the phonatory range of ex vivo rabbit larynges. It was found that increasing airflow and elongation affected subglottal pressure, fundamental frequency, sound pressure level (SPL) and the vibratory amplitude. In 2018, Döllinger et al. [39] performed the first systematic study on ex vivo rabbit larynges, analyzing HSI, audio and subglottal pressure data and the influence of glottal gap characteristics on the phonatory process, as such comprehensive analyses were missing in the literature. Significant influences of applied airflow and the vocal fold elongation level on vocal fold closure insufficiency were detected [40,41]. In the study of Döllinger et al. [39], only the periodic type 1 vibration data was analyzed. In the context of data collection, however, additional subharmonic type 2, as well as aperiodic type 3 vibration data were observed and recorded. This subharmonic type 2 and aperiodic type 3 vibration data are subject to systematic investigation in the present work. To the best of our knowledge, no ex vivo study has been previously reported that systematically investigated phonatory parameters from high-speed, audio and subglottal pressure data for disordered (subharmonic type 2 and aperiodic type 3) vocal fold oscillations. Therefore, the aim of this study is to close existing research gaps by (1) investigating the impact of glottis closure insufficiency on phonatory parameters also in the range of type 2 and type 3 vibrations; (2) compare the results with those of Döllinger et al. [39] who reported on periodic type 1 oscillations; (3) compare which of both vocal fold dynamics (subharmonic type 2 or aperiodic type 3) result in worse parameter values; (4) investigate whether the regularity (subharmonic type 2, aperiodic type 3 oscillations) or the glottis closure has a stronger impact on the fluid-structure-acoustic interaction (i.e., interaction of airflow-vocal folds-resulting sound) of the phonatory process.

Materials and Methods
The data acquisition was already previously described in detail [39]; hence the following is only a short overview of the experimental setup and data collection. For detailed information, we refer to the reference study on type 1 oscillations [39]. Data from 11 ex vivo rabbit larynges (New Zealand White, 4-5 kg body weight, ages 14-118 weeks) were used for this study. Since the rabbits were already sacrificed for another study with buprenorphine, no further approval of the ethics committee was required. Ethical acceptability was approved for the previous study (approval number 54-2532.1-54/12). In preparation for the experiments, the larynges were harvested from the sacrificed rabbits. Then the tissue just above and 30 mm below the larynges was surgically removed. To preserve the tissue characteristics, the prepared larynges were quickly frozen at −150 °C in liquid nitrogen and then stored at −80 °C [42]. Before the experiments, the larynges were slowly thawed at 6 °C in a refrigerator. They were then fixated on a 4 mm inside diameter stainless steel tube, which functioned as an artificial trachea. For fixation, a stainless adjustable steel ring was used to prevent air leakage. Rods and positioning screws were used to maintain stable larynx positions, see Figure 1 in [39]. At a distance of 100 mm below the larynx, a hole was drilled in the stainless-steel tube and the pressure sensor was placed there to measure the subglottal pressure. Figure 2 shows the experimental setup.
The subglottal pressure data were measured with an XCS-93-5PSISG pressure sensor (Kulite Semiconductor Products, Inc., Leonia, NJ, USA), which was positioned 100 mm below the larynx on the inside of the stainless-steel tube. The pressure sensor was connected to a PXIe-4330 bridge module (National Instruments, Austin, TX, USA).
The acoustic pressure data was measured using a 4189 1/2" free-field microphone (Brüel & Kjaer, 2850 Naerum, Denmark) mounted above the glottis at a distance of 200 mm and a tilt angle of 45°. The microphone output signal from a Nexus 2690 microphone was then further processed by a PXIe-4492 dynamic signal acquisition module (National Instruments, Austin, TX, USA). Both signals were resolved at 24 bits.
The vocal fold vibrations were recorded by a Phantom V2511 high-speed camera (Vision Research, Wayne, NJ, USA) at 8000 frames per second (fps) with a spatial resolution of 768 × 768 pixels. The videos were recorded at 16 bits. A Canon EF 180 mm 1: 3.5 L USM macro lens (Canon, Ōta, Tokyo, Japan) was used.
The subglottal pressure signal as well as the acoustic signal were synchronously sampled at a rate of f s = 96 kHz. For this purpose, a PXIe-6356 multifunctional data acquisition module (National Instruments, Austin, TX, USA) was used. The setup was controlled using the software LabView (National Instruments, Austin, TX, USA). A detailed description of the setup and its control is given in Birk et al. (2017) [43].
Different levels of vocal fold pre-stress were induced by three different weights applied anteriorly to the thyroid cartilage (w 1 = 1 g, W 2 = 2 g, W 3 = 5 g). The weights were sutured to the thyroid cartilage and used to shift it forward, thereby simulating contraction of the cricothyroid muscle in three different phonatory positions. An MF1 mass flow controller (MKS Instruments, Andover, MA, USA) powered by a PR4000B digital power supply (MKS Instruments, Andover, MA, USA) was used to generate an airflow passing through the artificial stainless-steel trachea and larynx. That air was humified with water vapor from a Neptune Heated Humidifier (Teleflex, Morrisville, NC, USA), and heated to 37 °C to simulate physiological in vivo conditions. The HSI technique allows the computation of quantitative, dynamic-based parameters. Its usefulness has been demonstrated in several previous studies [39,[43][44][45][46]. The HSI data analysis was performed using the in-house software Glottis-Analysis-Tools (GAT) (University Hospital Erlangen, Erlangen, Germany). The basis for all HSI analyses is the segmentation of the change in the area between the vocal folds over time (i.e., glottis area; given in pixels), which is referred to as the Glottal-Area-Waveform (GAW). A characteristic GAW with the corresponding glottal images is shown in Figure 3. The GAW is used to compute various phonation parameters that describe the phonatory process at the vocal fold level and to obtain information about vocal fold oscillations [47] such as periodicity, glottis closure and left-right symmetry of the vocal folds.
Each experimental run was recorded by the HSI camera for a length of 125ms with sustained phonation (i.e., >42 vibratory cycles, meeting the criteria >20 cycles) as suggested by [48] while the subglottal and audio data were recorded for 500 ms [39].
The computed parameters are subdivided into the groups "GAW parameters", "Aerodynamic parameters" and "Harmonic measures" according to their data sources; see Table 1.
In the study by Döllinger et al. [39], only periodic vocal fold vibrations containing exactly one harmonic were considered; i.e., these vibrations are considered normal [39]. In contrast, only (1) periodic vibrations with two harmonics and (2) aperiodic vocal fold vibrations are examined in this present study; i.e., these vibrations are considered as disordered.  Figure 6). In these figures, the audio signal was shifted by 0.58 ms to correct for the time delay of the acoustic signal; i.e., distance from the vocal folds to the microphone (200 mm) × acoustic wave propagation in air (343 m/s). Question 1: How does glottal closure insufficiency, expressed by GGI (Table 1), influence the fluid-structure-acoustic interaction within disordered vocal fold dynamics?
The GGI groups were chosen equivalent to Döllinger et al. [39] in order to ensure comparability of the two studies. The three GGI groups are categorized as follows: Best-GGI 1 ([0; 0.01], desired complete glottis closure during vibration), Medium-GGI 2 (]0.01, 0.4[, partial closure of the vocal folds), Worst-GGI 3 ([0.4; 1], no contact of the vocal folds during phonation). Based on the small group sizes, Kruskal-Wallis tests were applied for multiple group comparisons. For subsequent post hoc tests, Mann-Whitney-U tests with a Bonferroni correction (0.05/3 = 0.017) with a resulting significance level of p = 0.017 were used Question 2: a) Do the subharmonic dynamics influence the phonatory fluid-structureacoustic interaction process differently compared to aperiodic dynamics? b) Which of both dynamics result in worse phonatory parameter values? First, the data was examined for the presence of a normal distribution (Shapiro-Wilk). The group differences of the normally distributed data were tested for statistical significance with the parametric t-test, those of the not normally distributed data with the Mann-Whitney-Utest. All statistical analyses were performed using IBM SPSS Statistics 22 software (IBM, Amonk, NY, USA). Table 2 provides an overview of the various fundamental phonatory parameters such as fundamental frequency f 0 (Hz), subglottal pressure P S (Pa), glottal air flow (mL s −1 ), flow resistance R B (Pa s −1 ), and intensity SPL (dB) with the corresponding mean values, minimum and maximum values.

Fundamental Phonatory Parameters
In this study, data of 51 test runs were used. Of these, 35 test runs contained subharmonic (Group S ) and 16 test runs aperiodic oscillations (Group A ). As Table 3 shows, the subharmonic and aperiodic oscillations occurred at all airflow intensities (different airflow levels) and elongation levels (different weight levels), but tended to increase with higher air flow and higher elongation.

Influence, of the Glottal Gap on the Phonatory Process
Question 1: How does glottal closure insufficiency, expressed by GGI, influence the fluidstructure-acoustic interaction within disordered dynamics?
The results of the statistical tests are presented in Table 4. The group differences of all phonation parameters used were tested, which occurred firstly between the three GGI groups and secondly between the two vibration characteristics (Group S and Group A ). There were 8 out of 13 tests (61.5%) between all three GGI groups and 10 out of 24 post hoc tests (41.7%) that showed statistically significant differences. Statistically significant differences were found in 6 out of the 8 GAW parameters. Most differences in the direct comparison of two GGI groups were found between GGI 1 and GGI 2 with 4 statistically significant differences. Between GGI 1 and GGI 3 , there were 2 statistically significant group differences ( Table 4) and 2 of the p-values were slightly above the corrected p-value of 0.017; both with p-values of 0.018. Regarding aerodynamic parameters, the SPL values between the three GGI groups were statistically significantly. For the harmonic measures, CPP A was statistically significantly different between GGI 1 and GGI 3 .

Influence of the Vibrational Characteristics Subharmonic and Aperiodic on the Phonatory Process
Question 2: a) Do the subharmonic dynamics influence the fluid-structure-acoustic interaction differently compared to aperiodic dynamics? b) Which of both dynamics result in worse parameter values?
The parameters ALR, Stiffness, SPL and CPP A were normally distributed. All other parameters were not normally distributed, so the nonparametric Mann-Whitney-U-test was used for direct group comparison. The corresponding p-values are given in Table 4 in the last column. When comparing the mean values of aperiodic and subharmonic values, statistically significant differences were found in 4 of 8 GAW parameters (ALR, ASQ, CQ, PAI). Of the aerodynamic parameters, only the SPL values were statistically significantly different (p < 0.001). There were no statistically significant differences in the harmonic measures. Table 5 shows the means and standard deviations for all parameters of the three GGI groups (GGI 1 (N = 16), GGI 2 (N = 29), GGI 3 (N = 6)). The GAW parameters CQ and OQ increase steadily from GGI 1 to GGI 3 , while for the parameters ASI and PAI no clear trend is discernible. The remaining parameters decrease from GGI 1 to GGI 3 . With the exception of P S , the values of the other two aerodynamic parameters R B and SPL and those of the harmonic parameters CPP A and CPP P decrease from GGI 1 to GGI 3 .

Descriptive Statistics
The mean and standard deviations of all parameters for the two different vibrational characteristics are given in Table 6. From Group S to Group A , all GAW parameters except for ASQ, SQ and PAI decrease; see Table 6. Of the aerodynamic parameters, SPL and P S decrease, while R B increases from Group S to Group A . Both CPP A and CPP P decrease from Group S to Group A .

Discussion
A comparison of the computed parameters with other ex vivo rabbit studies has already been reported in Döllinger et al. [39], hence we refer to this work for the interested reader. The main focus of this work is the comparison of periodic type 1 vibrations from [39] with the type 2 and type 3 oscillations in this study, as well as on the comparison within the irregular vibration data.

Fundamental Phonatory Parameters
In comparison with the mean values of Döllinger et al. [39], it is noticeable that the fundamental frequency f 0 shows higher mean values (+8%) as do the minimum (+9%) and maximum values (+4%); see Table 2. This may be an indication for pathological voice, since Yamauchi et al. [58] found a relationship between incomplete glottis closure and an increase of the fundamental frequency, while investigating data of patients with vocal fold paralysis. Also, similar findings were reported by Wolfe et al. [59] who found a correlation between jitter and the degree of dysphonia by considering the mean fundamental frequency. Hence, an increased fundamental frequency may be seen as an indicator of a pathological or disordered voice [60]. Air flow strength here (120 mL/s) is close to the value of Döllinger et al. [39] of 117 mL/s, which was to be expected since the data were collected during the same series of experiments. Figure 7 shows the fundamental frequencies (GAW, audio and subglottal). This figure shows that the frequency of the GAW is the lowest and that of the subglottal and audio measurements tend to be higher. However, due to the aperiodicity of the signals, this tendency is not necessarily a causal interrelation. Hence, an interpretation of this effect is difficult.

Phonation Parameters
In comparison to Döllinger et al. [39] not all parameters were computed. The parameters Jitter, Shimmer [61] and Harmonic-To-Noise-Ratio (HNR) [62] are computed in the time domain of the signal and need quasi-periodicity of the signal [63]. Since the aperiodic data show reduced or no periodicity, it was not reasonable to compute and consider these parameters.

GAW Parameters-Eight
GAW parameters were considered. The ALR parameter differs statistically significantly between GGI 1 , 3 as well as between GGI 2 , 3 .
Higher ALR values indicate greater deformability and are desirable [39]. In this study, as well as in Döllinger et al. [39]. ALR decreases from GGI 1 to GGI 3 . Since GGI 3 has no vocal fold contact, a lower dynamical deformability and consequently a lower value of ALR is plausible compared to GGI 1 and GGI 2 . It is noticeable that the mean ALR for GGI 3 from this study is 37% lower compared to the normal ALR at GGI 3 in [39]; see Table 7. These worse results suggest less deformability for the disordered oscillation data. In comparison of the two vibrational characteristics, the values of the ALR parameters were statistically significantly worse for Group A compared with Group S ( Table 4). The difference between the two mean values was −55% (Table 6). Comparing the lowest mean ALR values of GGI (Table 5) and the vibrational characteristics (Table 6), mean ALR was considerably lower for GGI 3 at a mean value of 4.92, compared with the mean ALR value for Group A (8.14). This suggests that the combination of disordered oscillations (Group S or Group A ) and GGI 3 is particularly unfavorable with respect to the oscillatory characteristics of the vocal folds.
The two GAW parameters ASI and PAI, which reflect the dynamic left-right symmetry of the vocal folds, are associated with voice pathologies [58,64]. Surprisingly in this study, as well as in Döllinger et al. [39] the ASI and PAI values had better values for GGI 2 compared to GGI 1 and GGI 3 . Regarding the vibrational characteristics, it should be noted that the values both for ASI and PAI showed worse values for Group A compared to Group S ( Table  6). The PAI values were statistically significantly worse for the aperiodic vibrations (p = 0.009). This may be an indication that aperiodic oscillations are related to voice pathologies [58,64]. The ASQ values drop for GGI 1,2 from 0.60 to 0.49 and then increases to 0.52 for GGI 2 3. In comparison, in Döllinger et al. [39], the ASQ values decreased continuously from 0.62 to 0.59 with a glottal gap increase from GGI 1 to GGI 3 . Unsurprisingly, there are statistically significant differences in the CQ parameter between the GGI groups, because GGI 1 has a complete glottis closure and GGI 2,3 show no or only partial glottis closure. The difference in the CQ is also statistically significant between the two vibrational characteristics (Group S and Group A ). Regarding the GAW data, it is further noticeable that the parameters ASQ, CQ and SQ show strong fluctuations in the mean values of the three GGI groups in comparison to normal data (Table 7). These fluctuations without obvious trends lead to the assumption that these GAW parameters do not seem appropriate to indicate disordered voice.

Aerodynamic Parameters-
The energy transfer flow-tissue (R B ) between the two vibrational characteristics (Group S , A ) increased by 12% to 12363 (Pa s −1 ). In Döllinger et al. [39] R B decreased continuously for increasing glottal gap (GGI 1-3 ) by almost 50%. In contrast, here for GGI 1-2 R B decreased by 30% and then increased by 5% for GGI [2][3] . Compared to the data of Döllinger et al. [39] the value of R B for GGI 3 in this study was 42% higher; see Table 7. For GGI 1-2 our findings support other studies that higher energy transfer R B from the glottal flow to vocal folds yields an improved acoustic quality (i.e., higher CPP values) [39,60,65]. Surprisingly, we found contrary results for GGI [2][3] , that R B was increasing/improving while CPP decreased/deteriorated even further; see Table 5. This discrepancy could be due to the fact that none of the other studies were investigating data containing solely disordered oscillations. Further investigations on disordered dynamics are required.
The aerodynamic parameter SPL (dB) differs statistically significantly between the GGI groups, as well as between the two vibrational characteristics (Group S , Group A ); see Table  4. The intensity of the voice signal SPL (dB) decreases from Group S to Group A by 16% from 78.1 dB to 65.9 dB, indicating that aperiodic oscillations (Group A ) are worse compared to subharmonic oscillations (Group S ) in terms of efficiency/loudness of voice production. SPL also decreases from GGI 1 to GGI 3 by 22% from 76.7 to 59.5 dB (Table 5). This deterioration of SPL from GGI 1 to GGI 3 is stronger than for normal oscillations [39] from 79.1 to 69.4 (−12%). The general decrease in the intensity of the acoustic signal (SPL) may be explained by the fact that with a higher glottal gap index there is a limited oscillatory mobility of the vocal folds as well as an inadequate glottis closure and therefore sound is produced less efficiently.
It is noteworthy that for GGI 3 60% higher subglottal pressure than in Döllinger et al. [39] was needed to produce the intensity of 59.5 dB, which is 14% lower than in the data of Döllinger et al. [39]; see Table 7.
Various studies investigated the influence of high subglottal pressure and irregular oscillation (i.e., disordered; Group S and Group A ) in voice and rough phonation [28,66,67]. We cannot confirm these findings as the overall subglottal pressure regarding the irregular oscillations (Group S and Group A ) here (Table 2) is 8% lower (1324 Pa) than that of the normal oscillations of Döllinger et al. [39] (1436 Pa).
However, we did find an interrelationship between high subglottal pressure and glottal closure insufficiency. The subglottal pressure P S for GGI 1 , 2 was relatively similar comparing both studies while PS for GGI 3 in this study was 60% higher compared to the normal oscillation data [39]; see Table 7. We assume that the combination of strong glottis closure insufficiency (GGI 3 ) and disordered voices could be directly linked to high subglottal pressure. Unfortunately, previous studies on phonation data [28,66,67] did not examine the glottis closure insufficiency (i.e., GGI), which is why further investigations are needed to verify this assumption. Since the subglottal pressure data cannot be easily measured in humans, more ex vivo studies are hereby necessary.

Harmonic Measures-
The parameter CPP is currently one of the most applied parameters for characterizing dysphonia [68][69][70][71][72][73]. It is a measure of the voice quality that reflects the degree of occurring harmonics in a voice sample [57,74]. In addition, due to its calculation method, it is suitable for application to aperiodic signals, which is why it is of particular importance in the present study [75].
Just like in Döllinger et al. [39] the values of CPP A and CPP P decreased from GGI 1-3 , and thus became worse (Tables 5 and 7), as the CPP describes the quality of the voice signal [57,68,74,76,77]. Compared to Döllinger et al. [39] it is noticeable that the CPP values for the disordered GGI 1,2 are about 30% below the normal CPP values for GGI 1,2 , while the disordered CPP value for GGI 3 is about 45% below the normal CPP values for GGI 3 ; see Table 7. Meaning that these deteriorations of CPP for GGI 1-3 are considerably stronger for the disordered oscillations. These findings apply both to CPP A and CPP P . Deteriorations of CPP A and CPP P were also found from Group S to Group A , indicating again that aperiodic oscillations (Group A ) are worse for the acoustic quality compared to subharmonic oscillations (Group S ). This deterioration was however not as strong as for GGI [1][2][3] (Tables 5  and 6).
As Table 8 shows, GGI 3 is mainly composed of Group A . So, the poor CPP i values for GGI 3 may be based on this over proportional composition. However, since the mean CPP A (15.4 dB) and CPP P (16.0 dB) values of Group A ( Table 6) are above the CPP A (11.0 dB) and CPP P (14.0 dB) values of GGI 3 (Table 5), it is the combination of GGI 3 and irregularity that seems to be crucial for these poor CPP scores (i.e., reduced harmonics in the acoustic and subglottal signals).

Summary
Only five of the 13 investigated parameters show the same tendencies for both Group S to Group A and GGI 1 to GGI 3 ; see Tables 5 and 6. With increasing glottis closure insufficiency (i.e., GGI 1-3 ) nine of the 13 investigated phonatory parameters show similar tendencies compared to the periodic data of Döllinger et al. [39]. The worst mean values of the 4 parameters (ALR = 4.9, SPL = 59.5, CPP A = 11.0, CPP P = 14.0) that are directly linked to voice pathologies, voice disorders and dysphonia such as breathiness and roughness of voice [57,68,74,76,77] were found for GGI 3 . Therefore, we can confirm the findings of Döllinger et al. [39] that the more complete the glottis closure during phonation, the better the acoustic output. Table 7 shows the percentage change of the (disordered) phonatory parameters compared to (normal/non-disordered) Döllinger et al. [39]. Although GAW and aerodynamic parameters differed compared to [39], no obvious tendency was recognizable. A clear change was observable in the harmonic measures expressed by CPP A and CPP P . Both the quality of the subglottal pressure signal (CPP P ) and of the audio signal (CPP A ) were noticeably worse compared to [39]. This deterioration (decrease of CPP A,P ) became stronger with increasing glottis closure insufficiency (i.e., GGI 1-3 ), see Table 7.

Shortcomings
A limiting factor could be the length of time the larynges were frozen. The freezing, storing and thawing was performed according to the study of Chan and Titze [42]. Chan and Titze found that the vocal fold mucosa did not seem to change significantly after 24 h of storage in saline solution at room temperature, nor after one month of frozen storage following quick freezing. Therefore, their findings support the feasibility of using quick freezing to preserve laryngeal tissues for excised larynx experiments, as we performed. However, our larynges were frozen for a maximum of 13 months while the larynges of Chan and Titze were frozen for only one month. This longer time could have an effect on the tissue properties; however, there are no references to this assumption in the literature.

Conclusions
Since severe deterioration of the subglottal (CPP P ) and audio (CPP A ) signal quality occurred regardless which of the two irregular vibration types (subharmonic and aperiodic) were present, we suggest that these two parameters should be considered in medical diagnosis and treatment, alongside with glottal closure characteristics.
As Table 3 shows, disordered oscillations seemed to occur more often under high airflow levels (43%) and high elongation levels (55%). This may suggest that high or over stimulation of laryngeal parameters facilitates disordered phonatory behavior.
To further confirm these assumptions, we suggest further studies investigating disordered laryngeal dynamics with glottis closure insufficiency at different stimulation levels and their impact on the phonatory process, using ex vivo larynx experiments.    [57] Cepstral peak prominence Development of harmonics, the higher the better low: low periodicity of the acoustic signal high: high periodicity of the acoustic signal (computed from the audio "A" and the subglottal pressure "P" signal)  Table 3.
Frequencies of the test-runs divided into the different airflow levels and elongation levels.    Table 7.
Percentage change of this present data to normal data in corresponding GGI groups, Döllinger et al. (2018) [39].