Relative accuracy and reproducibility of regional MRI brain volumes for point-counting methods

Volumetric magnetic resonance imaging (MRI) methods for the measurement of various neuroanatomical regions are of great interest in studies of neuropsychiatric disorders. Both manual and semiautomated methods have been developed. Manual methods include tracing and point counting. Point counting methods are widely used in post-mortem and microscopy studies. Point counting has been well validated for these purposes. In this article, we report in a series of separate studies the accuracy and reproducibility of point-counting methods. Absolute accuracy was evaluated with a spherical phantom. Accuracy and time efficiency were subsequently assessed with an anatomically realistic phantom and various size grids. The point-counting method was also compared to a tracing method. Finally, the reproducibility of the point-counting method for the caudate and putamen was evaluated on four subjects in a test-retest experiment. These studies provide an estimate of the accuracy and time efficiency of point-counting methods. The test-retest reliability was also high for both caudate and putamen. Findings suggest that point counting is a reliable and efficient method for estimating volumes.