Abstract
1455
Objectives LV blood pool phase analysis has been investigated as a means of quantifying cardiac dyssynchrony, but a method for assessing the errors associated with the phase analysis parameters has yet to be investigated. In this abstract, a computational method of determining phase analysis errors is proposed. Ideally, the errors can be determined by repeated scanning of the patient and calculating the standard deviation of the measurements. In the proposed method, repeated scanning of the patient is approximated by applying a noise model to the measured data.
Methods Fifteen normal subjects were imaged using 24 gated blood pool scintigraphy. The time-activity curve (TAC) for each pixel in the LV was measured. Phase analysis was applied to the LV TACs to determine the standard deviation in phase values (phaseSD), synchrony, and entropy. The errors for these three parameters were then calculated using an iterative approach. For each iteration, Poisson noise was added to the LV TACs, and phaseSD, synchrony, and entropy were recalculated and averaged over all iterations. The error, for each parameter, was defined as the standard deviation over all iterations divided by the mean (expressed as a percentage). The effect of the number of iterations on the error and correlation with image noise was assessed.
Results For all parameters, a plot of the error vs. number of iterations showed a plateau at 500 iterations and above. For 500 iterations, the error values were not found to be correlated to image noise for any of the three parameters (r values all less than 0.2). This may be a result of the high image quality of the normal scans. At 500 iterations, the error values for phaseSD, synchrony, and entropy were found to be 6.5%, 1.7%, and 1.0% respectively.
Conclusions A method for determining phase analysis parameter errors has been developed.
- © 2009 by Society of Nuclear Medicine