Advanced Segmentation Techniques

Functional imaging with PET, SPECT, and/or dynamic MRI provides in vivo quantitative measurements of physiologic parameters of biochemical pathways and physiology in a noninvasive manner. A critical component is the extraction of physiological data, which requires accurate localization/segmentation of the appropriate ROIs. A common approach is to identify the anatomic structures by placing ROIs directly on the functional images, and the underlying tissue TACs are then extracted for subsequent analysis. This ROI analysis approach, although widely used in clinical and research settings, is operator-dependent and thus prone to reproducibility errors and it is also time-consuming. In addition, this approach is problematic when applied to small structures because of the PVEs due to finite spatial resolution of the imaging devices.

Methods discussed so far can be applied to almost all kinds of image segmentation problem because they do not require any model (i.e. model-free) that guides or constrains the segmentation process. However, segmenting structures of interest from functional images is difficult because of the imprecise anatomical information, the complexity and variability of anatomy shapes and sizes within and across individuals, and acquisition artifact, such as spatial aliasing, and insufficient temporal sampling, noise, and organ/patient movements. All these factors can hamper the boundary detection process and cause discontinuous or indistinguishable boundaries. Model-free approaches usually generate ambiguous segmentation results under these circumstances, and considerable amounts of human intervention are needed to resolve the ambiguity in segmentation. In this section, some advanced segmentation approaches are introduced, including

• model-based segmentation techniques that use analytical models to describe the shape of the underlying ROI,

• multimodal techniques that integrate information available from different imaging modalities for segmentation, or the image measurements are transformed and mapped to a standard template, and

• multivariate approaches are data-driven techniques in which the structures are identified and extracted based on the temporal information present in the dynamic images.

0 0

Post a comment