Frequency analysis of the sneeze caused by swine influenza virus strains: Automatic sneeze around-the-clock detection using a support vector machine

Abstract

Swine influenza, in particular, exhibits very fast transmission to piggeries or pig farms because it is a respiratory infection that can spread via air. Typical symptoms of swine influenza include sneezing. The pathogenicity of swine influenza depends on the specific virus strain. In previous researches, the number of sneezes was counted by a human inside a pigpen, only once-a-day, one-hour monitoring. In addition, this work conducted at the same time as nasal swabs. However, factors other than viral infection, such as environmentally derived stimuli like the existence of illumination and feeding practices can also induce sneezing in pigs. Instead, automatic measurements over a 24-h period can help rule out environmental causes other than viral strains and expedite infection diagnoses. The primary aim of this study was to investigate the relationships between the number of sneezes and different strains of the influenza virus in domestic pigs. We measured the number of sneezes constantly during a two-week infection experiment using an automatic sneeze detector implemented using a support vector machine. We analyzed the relationship between the number of sneezes and various factors in swine influenza infections. Prior to the experiment, we compared the sneeze-detection performances corresponding to each microphone position based on the acoustic features of sneezes. We compared the durations of sneezes and each other experimental factor per hour. The classification performance of the automatic sneeze detector was observed to be close to 100%, indicating that the automatic sneeze detector did not overfit the microphone data. We also observed a relationship between increased numbers of sneezes and other experimental factors. The number of sneezes increased following meals and infection check times in each group. At night, when there were few additional stimuli induced by factors associated with the infection experiment, we detected an infection three days after introducing it in the A/swine/Yokohama/aq114/2011(H3N2) and A/swine/Kagoshima/30/2015(H1N1) groups. This demonstrates that infection of certain swine influenza virus strains can be detected a few days after the infection by analyzing the frequency of induced sneezes.