Date of Award

Spring 1999

Document Type

Dissertation

Degree Name

Doctor of Engineering (DEng)

Abstract

Computer analysis of human respiratory sounds (phonopneumography) has been attempted since the 1960s with poor results. Steady research efforts by numerous researchers have failed to yield a practical diagnostic tool to replace the physician and the stethoscope. This has been the result of the high degree of variability in proposed analysis parameters from subject to subject.

The purpose of this research is to examine a new approach to phonopneumography by comparing multiple lung fields of the same subject for the same breath in order to develop the diagnostic parameters. This approach has the advantage of allowing the subject to provide the norm for comparison. Such an approach is possible because lung pathologies rarely affect both lungs in the same location in the same way, unless the condition is so severe as to be easily diagnosed without auscultation.

Two pediatric subjects were recorded in four locations on the anterior torso consisting of two pairs of homologous lung fields. These recordings were analyzed for their spectral content, and parameters relating to the power distribution were calculated. These parameters were compared to seek similarities between homologous segments for each subject. In addition, the researcher compared these parameters for each pair of segments in the same lung to seek an additional parameter that might prove useful for diagnosis. Finally, the parameters for each subject were compared to see if the similarity would extend beyond the individual subject.

A strong similarity was noted in homologous lung fields for each subject. Additionally, each subject showed a rising central frequency as the analysis moved to a higher location on the lung. No such similarity was noted when comparing the two subjects' data.

Share

COinS