An automated method for delineating a reference region using masked volumewise principal-component analysis in 11C-PIB PET

Abstract

Kinetic modeling using a reference region is a common method for the analysis of dynamic PET studies. Available methods for outlining regions of interest representing reference regions are usually time-consuming and difficult and tend to be subjective; therefore, MRI is used to help physicians and experts to define regions of interest with higher precision. The current work introduces a fast and automated method to delineate the reference region of images obtained from an N-methyl- 11C-2-(4′-methyla-minophenyl)- 6-hydroxy-benzothiazole ( 11C-PIB) PET study on Alzheimer disease patients and healthy controls using a newly introduced masked volumewise principal-component analysis. Methods: The analysis was performed on PET studies from 22 Alzheimer disease patients (baseline, follow-up, and test/retest studies) and 4 healthy controls, that is, a total of 26 individual scans. The second principal-component images, which illustrate the kinetic behavior of the tracer in gray matter of the cerebellar cortex, were used as input data for automatic delineation of the reference region. To study the variation associated with the manual and proposed automatic methods, we defined the reference region repeatedly. Results: As expected, the automatic method showed no variation whereas the manual method varied significantly on repetition. Furthermore, the automatic method was significantly faster, more robust, and less biased. Conclusion: The automatic method is helpful in the delineation of the reference region of 11C-PIB PET studies of the human brain and is much faster and more precise than manual delineation. Copyright © 2009 by the Society of Nuclear Medicine, Inc.