Prediction of forced expiratory volume in spirometric pulmonary function test using adaptive neuro fuzzy inference...

Spirometry is the most frequently performed clinical test to assess the dynamics of pulmonary function in human subjects. It measures airflow from fully inflated lungs through forced expiratory maneuver and generates large data set. However, these investigations often result in incomplete data sets due to the inability of the children and patients to perform this test. Hence, there is a requirement for prediction of significant parameters from the available incomplete data set. In this work, the results of model based prediction of two such significant parameters, Forced Expiratory Volume in one second (FEV1) and, Forced Expiratory Volume in six seconds (FEV6), are reported. The measured spirometric parameters are given as inputs to the Adaptive Neuro Fuzzy Inference System (ANFIS) which classifies data sets using fuzzy system based multilayer architecture. Triangular, Trapezoidal, Gaussian, Pi and Gbell membership functions are used to train and test the prediction process. The performance of the model is evaluated by computing their prediction error statistics of average value, standard deviation and root mean square. Results show that ANFIS model is capable of predicting FEV1 and FEV6 in both normal and abnormal subjects. Trapezoidal membership function predicted FEV1 with high precision and accuracy using a set of 21 rules. Similar prediction accuracy is observed in FEV6 using Gaussian membership function. Further, it is observed that prediction accuracy is found to be high for normal subjects with better correlation with measured values. It appears that this method is useful in enhancing diagnostic relevance of spirometric investigations in case of children and patients who are not able to perform the test as FEV1 and FEV6 are the useful indices to characterize pulmonary abnormalities.