Cough sound analysis can rapidly diagnose childhood pneumonia.

摘要: Pneumonia annually kills over 1,800,000 children throughout the world. The vast majority of these deaths occur in resource poor regions such as sub-Saharan Africa and remote Asia. Prompt diagnosis proper treatment are essential to prevent unnecessary deaths. reliable childhood pneumonia is fraught with difficulties arising from lack field-deployable imaging laboratory facilities well scarcity trained community healthcare workers. In this paper, we present a pioneering class technology addressing both problems. Our approach centred on automated analysis cough respiratory sounds, collected via microphones that do not require physical contact subjects. Cough cardinal symptom but current clinical routines used settings make use coughs beyond noting its existence screening-in criterion. We hypothesized carries vital information diagnose pneumonia, developed mathematical features pattern classifier system suited for task. sounds 91 patients suspected acute illness bronchiolitis asthma. Non-contact kept by patient's bedside were data acquisition. extracted non-Gaussianity Mel Cepstra them train Logistic Regression classifier. provided paediatric clinician gold standard validate our methods proposed paper could separate other diseases at sensitivity specificity 94 75% respectively, based parameters alone. inclusion simple measurements presence fever further increased performance. These results show indeed carry critical lower tract, can be pneumonia. performance method far superior those existing WHO algorithms resource-poor regions. To best knowledge, first attempt world humans using sound analysis. has potential revolutionize management