| A methodology which efficiently extracts characteristic features of the human thorax response subjected to a measured mechanical impulse (percussion) on the sternum manubrium for the purpose of disease detection is presented. For the first time medical percussion has been transformed into a standardized system for measurement and subsequent analysis. It has been found that the system displays Chaotic behaviour. A degree of time variance is present, caused in part by the inflation cycle (a variable which is hard to control in patients with impaired lung function), while non-stationarity exists due to the transient nature of the response. Extracting features using advanced digital processing techniques, we utilize a neural network to recognize disease behaviour. It is demonstrated both practically and statistically, that lung disease is detectable by this methodology.; Wavelet analysis and theory suggests the thorax to be composed of two resonanting systems, a lung component travelling slower than a second thorax shell component. The lung component alters in disease while the second thorax component in theory should not alter to the same degree. Analysis suggests that both are in non-linear coupled response and that interaction of these systems accounts for the overall system behaviour when the lung is in disease.; Evidence from this empirical study suggests that in some cases detection of abnormality was possible before X-ray radiographic features had appeared, supporting medical publications which suggest percussion could form the basis of a powerful diagnosis tool in early disease. |
