This study describes a non-invasive method for assessment of the lung transit time distribution of a tracer, using first-pass technetium-99m albumin angiocardiography and a model-free method of deconvolution. Ten patients received a first injection of 1 MBq kg-1 in the external jugular vein to position a gamma camera in the left anterior oblique position and two additional injections (5 MBq kg-1) to record first-pass angiocardiographic data. Right and left ventricular time-activity curves were derived from regions of interest every 0.5 s over a 1-min period. The left ventricular curve was deconvoluted by the right ventricular curve to obtain the lung transport function. The deconvolution procedure was based on a modified version of the Kalman filtering technique. The procedure was repeated at an interval of 30 min in eight patients. Two patients were re-examined up to 2 years later. Skewness, kurtosis and relative dispersion of the distributions did not change over time. We also found that the distribution, once normalized by its first moment, was independent of isolated changes in heart rate or cardiac output. Comparison of curve shapes at an interval of 30 min by point by point analysis demonstrated the reproducibility of the technique. We conclude that computation of the pulmonary transit time distribution of 99mTc-albumin from a standard angiocardiography procedure by model-free deconvolution is reliable and reproducible over time. We suggest that it may be a valuable tool for the non-invasive follow-up of the pulmonary circulation.