BACKGROUND: Noninvasive estimation of pulmonary artery pressure is an important component of echocardiographic studies. A number of methods are available for estimation of pulmenary pressure, each with varying degrees of reported accuracy. To assess accuracy and difficulties, noninvasive pulmonary artery pressure estimates were performed in infants and children with congenital heart diseases. METHODS: Noninvasive estimates from 8 methods were compared with catheterization measurements. Systolic pressure was estimated by the Burstin method and from perak tricuspid regurgitation velocity, and also from systolic pressure gradients through the VSD(ventricular septal defect) and PDA(patent ductus arteriosus). Mean pressure was estimated by acceleration time divided by ejection time measured from Koppler spectrum obtained at the right ventricular out flow tract. Diastolic pressure was estimated from pulmonary regurgitation velocity spentrum at end-diastolic, and also from diastolic pressure gradient through the patent ductus arteriosus. RESULTS: IN systolic pressure, Burstin and tricuspid regurgitation velocities estimates correlated significantly(r=0.92, 0.90 respectively), whereas VSD and PDA estimates correlated less well with catheterization estimates(r=0.83, 0.65 respectively). The mean pressure, measured from RVOT(right ventricular outflow tract) Doppler spectrum corresponded well with catheterization pressure(r=0.89), whereas those obtained from the main pulmonary artery correlated less well(r=0.74). The diastolic pressure estimates from pulmonary regurgitation velocity spectrum, revealed good correlation(r=0.79), but those from diastolic Doppler spectrum at PDA correlated less well with catheterization estimates(r=0.63). CONCLUSION: All of eight Doppler echocardiographic methods seemed to be easily performable for estimation of pulmonary artery pressure. But, the degree of accuracy was variable. Because a pressure estimante from only a single method may be in error, care should be taken in combining use of other(one or two) methods.