Constructing Atlases and Measuring Structural Brain Changes via Diffusion Tensor Imaging
Susumu Mori
Johns Hopkins University
Magnetic Resonance Imaging (MRI) is widely recognized as one of the most versatile radiological techniques for noninvasive study of the human brain. Its versatility arises from its ability to generate multiple patterns of image contrast, depending on the data acquisition techniques employed. Each contrast mechanism is based on different physical and chemical properties of water molecules, and thus each contrast pattern often reflects distinct physiological and/or anatomical properties of the brain.
In the late 1980s and early 1990s, a new MRI contrast scheme known as diffusion imaging was introduced. This contrast mechanism is unique because it is sensitive to the structural orientation of axons, which cannot be studied by other radiological techniques. Diffusion imaging uses the water diffusion process as a probe to investigate brain axonal organization, based on the principle that water tends to diffuse preferentially along axonal fibers. From its inception, it was clear that this technique would provide valuable new information about white matter anatomy. However, it remains mathematically challenging to describe the brain’s highly complex neuroanatomy solely through the water diffusion process.
Currently, one of the most widely accepted approaches is based on a tensor model, commonly referred to as diffusion tensor imaging (DTI). In the first part of this lecture, I review the fundamentals of diffusion measurement and data processing within the tensor framework, with special emphasis on understanding the processing pipeline from raw MR images to computed fiber orientation maps.
In the second part, I introduce alternative approaches for modeling the diffusion process and inferring the underlying anatomy. In the third part, I discuss strategies for applying computational neuroanatomy methods to white matter analysis using diffusion tensor imaging, along with relevant biological applications.
