Research Article: An advanced recording and analysis system for the differentiation of guinea pig cough responses to citric acid and prostaglandin E2 in real time

Date Published: May 22, 2019

Publisher: Public Library of Science

Author(s): Jianguo Zhuang, Lei Zhao, Xiuping Gao, Fadi Xu, Yu Ru Kou.


Cough number and/or sound have been used to assess cough sensitivity/intensity and to discriminate cough patterns in clinical settings. However, to date, only manual counting of cough number in an offline manner is applied in animal cough studies, which diminishes the efficiency of cough identification and hinders the diagnostic discrimination of cough patterns, especially in animals with pulmonary diseases. This study aims to validate a novel recording/analysis system by which cough numbers are automatically counted and cough patterns are comprehensively differentiated in real time. The experiment was carried out in conscious guinea pigs exposed to aerosolized citric acid (CA, 150 mM) and prostaglandin E2 (PGE2, 0.43 mM). Animal body posture (video), respiratory flow, and cough acoustics (audio) were simultaneously monitored and recorded. Cough number was counted automatically, and cough sound parameters including waveform, duration, power spectral density, spectrogram, and intensity, were analyzed in real time. Our results showed that CA- and PGE2-evoked coughs had the same cough numbers but completely different patterns [individual coughs vs. bout(s) of coughs]. Compared to CA-evoked coughs, PGE2-evoked coughs possess a longer latency, higher cough rate (coughs/min), shorter cough sound duration, lower cough sound intensity, and distinct cough waveforms and spectrograms. A few mucus- and wheeze-like coughs were noted in response to CA but not to PGE2. In conclusion, our recording/analysis system is capable of automatically counting the cough number and successfully differentiating the cough pattern by using valuable cough sound indexes in real time. Our system enhances the objectivity, accuracy, and efficiency of cough identification and count, improves the intensity evaluation, and offers ability for pattern discrimination compared to traditional types of cough identification. Importantly, this approach is beneficial for assessing the efficacy of putative antitussive drugs in animals without or with pulmonary diseases, particularly in cases without significant change in cough number.

Partial Text

Cough is an important respiratory defense mechanism [1, 2] and one of the most common symptoms in clinical complaints [3]. Both cough number and sound have been clinically used to evaluate cough sensitivity and intensity [4–13]. In addition, the analysis of cough acoustic differences (frequency, duration and amplitude) has been applied to discriminate cough patterns, especially in diagnosis of human pulmonary diseases and guidance of the relevant therapies [12, 14–18]. Although the animal cough model has been widely employed in cough pathophysiology and therapy, the current recording/analysis system applied in the model obviously has an unmet requirement for the sufficient determination of cough severity and discrimination of cough pattern.

A typical cough response in animals has been characterized by the simultaneous appearance of remarkable and unique changes in the respiratory flow, sound, and body posture (three key signals) [24, 25, 27, 28]. However, the disadvantages of the previous recording system are obvious in these studies. First, though respiratory flow and sound were simultaneously recorded, the animal body posture was watched by observer(s) during the experiment without video recording. Because the changes of these variables are transient, matching them in a real-time manner is difficult during the experiment, which could lead to miscounting of cough numbers. Second, discrimination of cough from sneeze was manually counted by playback of the recordings (offline) in which the synchronized signals of the body posture were lacking, which markedly diminishes the accuracy and efficiency of identifying a cough. Third, cough sound duration and intensity synchronized with respiratory signals were not previously quantitatively analyzed in animal cough models, while it has been extensively applied in clinical settings [4–13]. Fourth, analysis of sound waveforms and spectrograms enables discrimination of productive and non-productive cough sounds in humans [31, 32]; however, such analysis has not been investigated in animals. Our recording/analysis system is novel because all the three key signals are simultaneously monitored and recorded, and importantly, there is a very high accuracy in automatically counting the coughs with louder sound (accuracy = 99% in CA-evoked coughs (see Fig 2 and S1 Video). Compared to CA, the accuracy of automatic counting coughs in response to PGE2 is much lower (~35%) owing to no or weak cough sound. Airflow changes during the PGE2-evoked cough are uniquely characterized by appearance of a series of large expiration in a bout manner (Fig 4). Thus, using this airflow characteristic may be another way to define Type II cough. Importantly, our recording/analysis system is capable of analyzing the acoustic signals (waveform, spectrogram, CSII, and CSD in real-time; and PSD distribution and PCSI offline). The acoustic signals of cough can be applied to easily distinguish cough from other expiratory activities that lack cough sound. An expiration reflex, sigh, and/or sniffing also contain a relatively large expiration [52, 53]. Though these expiratory activities exist in conscious guinea pigs, they lack the cough sound [52]. Therefore, our recording/analysis system provides the objectivity, accuracy and efficiency of counting cough numbers, offers the ability to differentiate cough patterns, and adds the cough sound intensity to assist in the determination of cough intensity.

Cough is the most common symptom in clinical settings (accounting for one third of pulmonologist consultations) [3, 71, 72]. Changes of the cough sound may indicate the effectiveness of therapy or the progress of disease. Animal cough models have been extensively used to develop new antitussive drugs. This study establishes a novel system by which all signals including cough-relayed body posture, respiratory flow, cough appearance and acoustics are simultaneously recorded and analyzed in real-time. This system provides the objectivity and accuracy of counting cough numbers and gains the ability for differentiating cough pattern and determining cough intensity. Further study is required to validate the applicability of our system to discriminate cough patterns in animals with pulmonary diseases, such as idiopathic pulmonary fibrosis, COPD, or asthma.