Implementation and Validation of a Fully Automatic System... : Journal of Computer Assisted Tomography (original) (raw)

Image Processing

Implementation and Validation of a Fully Automatic System for Intra- and Interindividual Registration of PET Brain Scans

From the Uppsala University PET Centre, Department of Radiation Sciences, and Subfemtomole Biorecognition Project (J. L. R. Andersson), Uppsala, and Centre for Image Analysis, Uppsala, and Departments of Clinical Neurophysiology and Neuroradiology, Karolinska Institute/Hospital, Stockholm (L. Thurfjell), Sweden.

Address correspondence and reprint requests to Dr. J. L. R. Andersson at Uppsala University PET Centre, S-751 85 Uppsala, Sweden.

Abstract

Purpose:

Stereotactic coordinate spaces and methods to adapt subjects to that space are required when performing averaging of functional studies across subjects.

Methods:

A rapid and fully automatic method to perform intersubject registration and adaptation to a previously defined coordinate space has been developed and implemented. The implementation has been performed within an existing software developed to facilitate manual registration and adaptation, thus offering a versatile combination of automatic and manual tools. Furthermore, a novel measure, based on the _F_-statistic for intersubject (block) differences, for the assessment of intersubject goodness of fit was suggested and validated.

Results:

The intra- and intersubject registration was validated by its application to data from six human subjects participating in an activation study. The registration was performed both manually and automatically, and the results indicated that the automatic method performed at least as well as the manual. The block _F_-statistic was lower for the automatic method, and the _z_-scores were not significantly different for the methods. The localization of activated regions showed good agreement and differed by an average of 6 mm between the methods.

Conclusion:

It is concluded that the suggested method is a valuable alternative to the current manual approach.