Spirometry is the most common pulmonary function test. It provides useful information for early detection of respiratory system abnormalities. While decision support systems use normally calculated parameters such as FEV1, FVC, and FEV1% to diagnose the pattern of respiratory system diseases, expert physicians pay close attention to the pattern of the flow-volume curve as well. Fisher discriminant analysis shows that coefficients of a simple polynomial function fitted to the curve, can capture the information about the disease patterns much better than the familiar single point parameters. A neural network then can classify the abnormality pattern as restrictive, obstructive, mixed, or normal. Using the data from 205 adult volunteers, total accuracy, sensitivity and specificity for four categories are 97.6%, 97.5% and 98.8% respectively.