Variations on calculating left-ventricular volume with the radionuclide count-based method

Various methods for the calculation of left-ventricular volume by the count-based method utilizing red-blood-cell labeling with 99mTc and a parallel-hole collimator are evaluated. Attenuation correction, linked to an additional left posterior oblique view, is utilized for all 26 patients. We examine (1) two methods of calculating depth, (2) the use of a pair of attenuation coefficients, (3) the optimization of attenuation coefficients, and (4) the employment of an automated program for expansion of the region of interest. The standard error of the estimate (SEE) from the correlation of the radionuclide volumes with the contrast-angiography volumes, and the root-mean-square difference between the two volume sets at the minimum SEE are computed. It is found that optimizing a single linear attenuation coefficient assumed for attenuation correction best reduces the value of the SEE. The average of the optimum value from the end-diastolic data and that from the end-systolic data is 0.11 cm-1. This value agrees with the mean minus one standard deviation value determined independently from computed tomography scans (0.13-0.02 cm-1). It is also found that expansion of the region of interest beyond the second-derivative edge with an automated program, in order to correctly include more counts, does not lower the SEE as hoped. This result is in contrast to the results of others with different data and a manual method. Possible causes for the difference are given.