Detection of mouse cough based on sound monitoring and respiratory airflow waveforms

Abstract

Detection for cough in mice has never yielded clearly audible sounds, so there is still a great deal of debates as to whether mice can cough in response to tussive stimuli. Here we introduce an approach for detection of mouse cough based on sound monitoring and airflow signals. 40 Female BALB/c mice were pretreated with normal saline, codeine, capasazepine or desensitized with capsaicin. Single mouse was put in a plethysmograph, exposed to aerosolized 100 µmol/L capsaicin for 3 min, followed by continuous observation for 3 min. Airflow signals of total 6 min were recorded and analyzed to detect coughs. Simultaneously, mouse cough sounds were sensed by a mini-microphone, monitored manually by an operator. When manual and automatic detection coincided, the cough was positively identified. Sound and sound waveforms were also recorded and filtered for further analysis. Body movements were observed by operator. Manual versus automated counts were compared. Seven types of airflow signals were identified by integrating manual and automated monitoring. Observation of mouse movements and analysis of sound waveforms alone did not produce meaningful data. Mouse cough numbers decreased significantly after all above drugs treatment. The Bland-Altman and consistency analysis between automatic and manual counts was 0.968 and 0.956. The study suggests that the mouse is able to present with cough, which could be detected by sound monitoring and respiratory airflow waveform changes.

Figure 1. Mouse cough detection equipments.

(1) Bias flow generator; (2) Desiccant; (3) Nebuliser controller; (4) Nebuliser; (5) Plethysmograph; (6) Amplifier; (7) Speakers; (8) Monitor display.

Figure 2. Seven types of mice respiratory waveforms: (1) cough; (2) sneeze; (3) eupnea; (4) tachypnea; (5) breath-holding; (6) deep inspiration; (7) head-twitch.

Figure 3. Comparison of automated and manual mouse cough counts.

(A) Bland-Altman analysis. (B) Consistency analysis. Round circle represents automated cough counts; triangle indicates manual cough counts. The intra-class correlation coefficient (ICC) was 0.956 (95% confidence interval: 0.911∼0.978).

Figure 4. Automated analysis of mouse cough waveforms.

The superior blue part indicates mice cough respiratory waveform and the inferior part shows airflow rate slope replot. Coughs were judged based onthe compression threshold, slope threshold and expiratory timephase.

Figure 5. Recorded and amplified sound waveforms according to mouse cough and noises.

(A) The upper panel shows sound waveforms in green, and the bottom one shows air flow signals in blue. Cough signals defined by Finepointe software were displayed white color. Manual defined coughs were marked with “cough” and red hollow dots on the screen. Coughs were pointed with red arrows. According to the cough signals, cough sound waveforms were certified. Noise 1 pointed with yellow arrow was head-twitch sound. Noise 2 was the sound of knocking at chamber wall with no abnormal respiratory signals. (B) Three different types of amplified cough sound waveforms distinguished by Cooledit software. (C) Two types of noise sound waveforms, which are different in shape, but similar to cough ones.

Figure 6. Deep inspiration could be distinguished from coughs through the shape of respiratory signals.

Deep inspiration with air flow signals that appeared wider during the early phase and became narrowed later, in contrast to those produced by coughs. Head-twitch, accompanied with a cough-like sound but producing inverted V-shaped air flow signals that was readily distinguishable from coughs. A: Cough; B: head-twitch; C: deep inspiration; D: cough.