Atlas-assisted localization analysis of functional images

Abstract

Objective: This paper introduces a method for localization analysis of functional images assisted by a brain atlas. The usefulness of the system developed, based on this method, is analyzed for human brain mapping and neuroradiology. Materials and methods: We use an enhanced and extended electronic Talairach–Tournoux brain atlas, containing segmented and labeled subcortical structures, Brodmann’s areas, and gyri. The brain atlas serves as a tool for anatomy referencing, segmentation, labeling, registration, and providing 3D anatomical relationships. The process of localization analysis is decomposed into five steps: data loading, feature extraction, data normalization, identification and editing of loci, and getting labels and values. This analysis is supported by near real-time data-to-atlas warping based on the Talairach transformation. Metanalysis is enabled by merging the current and external lists of activation loci. Results: We have designed, developed, tested, and deployed a commercial system for atlas-assisted localization analysis of functional images. This is the first system where an electronic version of the Talairach–Tournoux brain atlas is used interactively for analysis of functional images. This system runs on personal computers and provides functions for a rapid normalization of anatomical and functional volumetric data, data segmentation and labeling, readout of Talairach coordinates, and data display. It also is empowered with several unique features including: interactive warping facilitating fine tuning of the data-to-atlas fit, a backtracking mechanism to compensate for missing landmarks and enhancing the outcome of the overall process of data analysis, navigation on the triplanar formed by the data and the atlas, multiple-images-in-one display with atlas-anatomy-function blending, a fast locus-controlled generation of results, editing of loci, multiple label display, and saving and reading of loci. The system normalizes a single image in near real-time (0.7 s), so analysis of anatomical and functional datasets can be done on-the-fly regardless of the number of slices. The same task performed by the state-of-the-art non-linear registration methods may require up to several days. Conclusions: The system is a useful tool for atlas-assisted localization analysis and a helpful adjunct to function/location metanalysis in human brain mapping research. It is also a step forward in bringing the atlas and the clinical data together within a practical and powerful solution that is fast and flexible, yet low-cost and affordable.