RT Journal Article
SR Electronic
T1 Data reduction for large scale cough studies using distribution of audio frequency content
JF European Respiratory Journal
JO Eur Respir J
FD European Respiratory Society
SP p4027
VO 38
IS Suppl 55
A1 Antony Barton
A1 Patrick Gaydecki
A1 Kimberley Holt
A1 Jacky Smith
YR 2011
UL http://erj.ersjournals.com/content/38/Suppl_55/p4027.abstract
AB Background: Recent studies have suggested that the objective quantification of coughing from sound recordings provides novel insights into the mechanisms underlying cough and the efficacy of therapies. However, reliable methods for minimisation of sound data are required to improve the feasibility of processing many and large patient data records for large scale studies of cough treatments for both manual and potential automatic cough counting.Aim: To determine if a developed system can identify periods of inactivity in sound recordings to significantly reduce record length without degrading data (i.e. inadvertent removal of cough sounds), referred to hereafter as destructiveness.Methods: Inactive periods of audio are identified by measuring the median audio frequency within small segments of recordings and removing those below a selected threshold. 200 randomly selected 15 minute periods known to contain cough, from 20 patients [healthy (5), COPD (5), asthma (5) and chronic cough (5); male (10)] were used, each recorded for 24hrs. To measure destructiveness, both the audio kept and removed by the algorithm were analysed by trained cough counters and compared to counts for the original files. Finally, the efficacy of the algorithm was determined by the reduction in record length achieved across all of the patient data.Results: The average resultant file size was 6.04% (54.4s) of the original (median 13.9s, iqr 56.4s) and the system erroneously removed 1.6% of coughs at a rate of 0.62 coughs h-1.Conclusions: The system has shown to be reliable for use in cough monitoring as an excellent means of removing large sections of audio and profoundly improving the efficiency of manual cough counting.