Pattern-character analysis of flow volume-curve for diagnosis of pulmonary obstruction

This study presents a different approach of the maximum expiratory flow volume curve analysis. Instead of computing only its classical parameters, we are interested in the perceptible components it raises up, that is its pattern. Patterns have been quantified and classified with algorithms used in pattern recognition. A criterion of likeness and a criterion of proximity of each character curve with a standard pattern are calculated. It permits to associate a character and a normal or a pathological standard pattern. The automata which has been designed for automatic and real-time acquisition and computation of the test is described. Results on populations of children and obstructive patients are then presented. It lets us assess the ability of the system for early bronchial diagnosis.