Publications
Text queries can be conducted by Author, Title, or Keyword.
Towards a metabolic image of function
Source: Function Bull Clin Neurosci 1988;53:68-72.
Author: Collins RC, Toga AW.
Abstract:
The science of neuroimaging is still in its infancy. Where is it going? Imagine this: A 45 year old man complains to his physician of difficulty with memory. A neurological exam, neuropsychological testing, EEG, and MRI are within normal limits. The question is this: Can we study functional activity within the circuits of memory to determine abnormality? The answer at present is no. To achieve this goal we must build a six-dimensional brain: the three dimensions of space, the fourth dimension of function, the fifth dimension of time, and the sixth dimension of statistical significance. We must be able to test statistically whether the timing and intensity of mental activity within a prescribed pattern of circuits is normal or abnormal. Research in progress in our laboratories is devoted to defining the limits of the metabolic image. There are four interrelated projects: spatial resolution, temporal resolution, biochemical process, and data representation. Spatial resolution in a metabolic image in defined primarily by the energy of the isotope and secondarily by methods of detection. Tritium isotopes such as 3H-deoxyglucose can be used to study cellular metabolism. Carbon 14 can be used to study patterns of activiyt within histological fields. Emulsion and film autoradiography give sharp images since they are applied directly to the tissue. The practical range of resolution for 14C is approximately 100 to 200 um. Positrons are detected by coincident counting of annihilation radiation by cameras at a distance from a point source. The practical resolution of new generation PET scanning is approximately 6 mm. Basic functional units are smaller than this. Columns in cortex identified by electrode recordings or by special stains are approximately 300-500 um in diameter. This has proved constant for all mammals including man. Due to the energy of positrons it is not possible to resolve activity in cortical columns unless they are separated by several centimeters. Data representation for metabolic images poses a challenging problem. The goal is to display and measure changes in functional activity within the 3D structure of brain wherever there are meaningful changes. The problem can be broken down into 2 major components. First, the geometry of the structure itself must be established. This means reconstructing a 3D brain from serial, 3D sections. Reestablishing the relationship between sections requires realignment. Algorithms for alignment have used fiducials, rigid anatomic landmarks or engineering techniques such as cross correlation principal axis, and centroid of the mass. The next problem is to visualize the surface. Since the process of serial sectioning destroys all geometric data along the axis perpendicular to the cutting plane, this surface must be synthesized. A wire frame model can be built that spans the surface of cortex and deep structures between each section. These wire-frame models may then be used to synthesize structural surfaces as they orignally existed. Surface rendering algorithms, which provide 3D, depth cueing throught the use of shading, perspective transformations, and other computer graphic techniques, make visualizing the original brain possible. The second major component is visualizing functional data. Each anatomical point in the 3D brain corresponds to a numeric value of glucose metabolism or cerebral blood flow. These rates can be visualized by using the dimension of color. If the computer system now understands the geometry of the structure of the brain, functional anatomy can be superimposed onto this structural framework. The trick is to display these rates without allowing the anatomic data to obstruct the view. This can be achieved by taking mathematical cuts into the brain or by rendering the surface transluscent to look into its depth. Experiments in neuroimaging have indicated how the 4D of function can be displayed quantitavely and comprehensively within the 3D of anatomy. Dimensions of time and statistical significance will require comparisons among different sets of animals. The methods to achieve this are at hand.
Source: Function Bull Clin Neurosci 1988;53:68-72.
Author: Collins RC, Toga AW.
Abstract:
The science of neuroimaging is still in its infancy. Where is it going? Imagine this: A 45 year old man complains to his physician of difficulty with memory. A neurological exam, neuropsychological testing, EEG, and MRI are within normal limits. The question is this: Can we study functional activity within the circuits of memory to determine abnormality? The answer at present is no. To achieve this goal we must build a six-dimensional brain: the three dimensions of space, the fourth dimension of function, the fifth dimension of time, and the sixth dimension of statistical significance. We must be able to test statistically whether the timing and intensity of mental activity within a prescribed pattern of circuits is normal or abnormal. Research in progress in our laboratories is devoted to defining the limits of the metabolic image. There are four interrelated projects: spatial resolution, temporal resolution, biochemical process, and data representation. Spatial resolution in a metabolic image in defined primarily by the energy of the isotope and secondarily by methods of detection. Tritium isotopes such as 3H-deoxyglucose can be used to study cellular metabolism. Carbon 14 can be used to study patterns of activiyt within histological fields. Emulsion and film autoradiography give sharp images since they are applied directly to the tissue. The practical range of resolution for 14C is approximately 100 to 200 um. Positrons are detected by coincident counting of annihilation radiation by cameras at a distance from a point source. The practical resolution of new generation PET scanning is approximately 6 mm. Basic functional units are smaller than this. Columns in cortex identified by electrode recordings or by special stains are approximately 300-500 um in diameter. This has proved constant for all mammals including man. Due to the energy of positrons it is not possible to resolve activity in cortical columns unless they are separated by several centimeters. Data representation for metabolic images poses a challenging problem. The goal is to display and measure changes in functional activity within the 3D structure of brain wherever there are meaningful changes. The problem can be broken down into 2 major components. First, the geometry of the structure itself must be established. This means reconstructing a 3D brain from serial, 3D sections. Reestablishing the relationship between sections requires realignment. Algorithms for alignment have used fiducials, rigid anatomic landmarks or engineering techniques such as cross correlation principal axis, and centroid of the mass. The next problem is to visualize the surface. Since the process of serial sectioning destroys all geometric data along the axis perpendicular to the cutting plane, this surface must be synthesized. A wire frame model can be built that spans the surface of cortex and deep structures between each section. These wire-frame models may then be used to synthesize structural surfaces as they orignally existed. Surface rendering algorithms, which provide 3D, depth cueing throught the use of shading, perspective transformations, and other computer graphic techniques, make visualizing the original brain possible. The second major component is visualizing functional data. Each anatomical point in the 3D brain corresponds to a numeric value of glucose metabolism or cerebral blood flow. These rates can be visualized by using the dimension of color. If the computer system now understands the geometry of the structure of the brain, functional anatomy can be superimposed onto this structural framework. The trick is to display these rates without allowing the anatomic data to obstruct the view. This can be achieved by taking mathematical cuts into the brain or by rendering the surface transluscent to look into its depth. Experiments in neuroimaging have indicated how the 4D of function can be displayed quantitavely and comprehensively within the 3D of anatomy. Dimensions of time and statistical significance will require comparisons among different sets of animals. The methods to achieve this are at hand.