Use of factor analysis in the evaluation of left to right cardiac shunts

We have compared two methods of data processing for the quantitation of left-to-right cardiac shunts using first-pass radionuclide angiography. These two methods are used for curve generation in the deconvolution analysis. The standard method involves manual definition of regions of interest. A newer method--factor analysis--provides automatic curve generation and is therefore more operator-independent. Both techniques yield curves of the venous input, lung, and background. The venous input curve is deconvolved by the lung curve, and the resultant unit impulse response is fitted by the gamma variate method to quantitate the left-to-right shunt fraction. Both techniques--factor analysis and regions of interest (ROIs)--separated the shunt patients (n = 16) from the control subjects (n = 20) with a p less than 0.001. There was less interobserver variability with the curves obtained by factor analysis than with those obtained by regions of interest. The coefficient of correlation of factor analysis results with oximetry, r, was 0.90. High sensitivity and specificity, each 94%, was achieved with curves generated by factor analysis. Time vs. activity curves generated by ROIs can achieve high sensitivity and low specificity, or vice versa, depending on the cutoff level defined for separation of the left-to-right shunt patients from the control group.