Abstract
Maximum expiratory flow-volume(MEFV)curve convey vital information about the condition of respiratory system and can be used for assessing the degree of lung function impairment. A novel method of diagnosis of respiratory system based on model parameters extracted from the MEFV curve is presented in this paper. The model is established using the lung pressure-volume relationship, and the dependence of airway resistance on lung volume and airflow. Genetic algorithm based optimisation technique was implemented to determine the model parameters which best match the MEFV curves. The parameters were determined for normal subjects as well as patients having chronic obstructive pulmonary disease(COPD). MEFV curves recorded from the subjects are compared with the waveforms obtained using the simulated model and they show a close fit. Usefulness of the model parameters is examined using artificial neural network with the modelparameters as inputs and the condition of the subjects, normal or having any of the respiratory diseases, as outputs. Backpropagation algorithm incorporating Levenbury-Marquardt optimisation technique was used for training the neural network The trained neural network was verified using test data and it was observed that the proposed technique provides high percentage success in diagnosis.
