First published in Personal Computer World, May 1996.
THE HUMAN BRAIN is the most complex object in the known universe. With recent developments in scanning techniques coupled with the visualization power of computer graphics, researchers are able to see inside the living brain as never before.
Dissect an arm and you can distinguish skin, bone, cartilege, muscle fibres and veins; dissect a brain and you find a fairly uniform substance. Where is the centre of passion, or the 'creativity' organ? In the seventies, it was suggested (mainly by computer scientists) that the substance of the human brain was directly analogous to the hardware of a computer; and that our thoughts, feelings and desires were the software and data structures.
At first glance, it's a compelling idea. But, on reflection, it explains without *really explaining*, and so it's bad science. In the eighties, this analogy lost favour, and was superseded by a more 'holistic' approach, which stressed the anatomy and biochemistry of the brain as the only necessary explanation of 'us'. And with this shift of focus came the need for analysis and visualization at an unprecedented scale of complexity.
There are two fundamental problems in understanding structures as complex as the human brain. First, how to obtain the data; second, how to visualize it. Traditionally, understanding human tissue has involved dissection and staining, then photography or drawing. These are essentially two-dimensional techniques, and they are invasive: to study the tissue, you have to kill it. When human brains are concerned, this is something of a drawback.
With the development of non-invasive scanning techniques such as Magnetic Resonance Imaging (MRI), Computerised Tomography (CT) and Positron Excitation Tomography (PET), it is now possible to capture the structure of organs without destroying them. These methods map two-dimensional cross-sections of tissue, which can then be combined into three-dimensional datasets. The breakthrough in visualizing this data came with the technique of 'volume rendering', where stacks of 2D slices are rendered as collections of 'voxels', the three-dimensional analogue of two-dimensional pixels.
With volume rendering, the structure of an object is represented as a truly three-dimensional data set, which can be processed with filtering algorithms, to identify and extract 3D structures and surfaces. This technique is already yielding remarkable results. The Voxel-Man project from the University of Hamburg, for example, contains volumetric data of the brain, skull, heart and abdomen, and enables researchers to perform 'virtual dissections' by slicing the data set with arbitrary planes, to reveal the internal structure.
Another ambitious undertaking is the Visible Human Project from the United States National Library of Medicine in conjunction with the University of Colorado. Here data for a complete male cadaver is already available on the Internet, as raw images gathered from CT, MRI and cryosections. The data set is huge: each image slice is 7 megabytes, and there are 1,871 of them. That's 15 gigabytes.
The future of computer graphical neuroimaging is exciting. With the advent of volume rendering and stereoscopic viewing, researchers have access for the first time to true three dimensional computer models of living tissue. But there are major problems still to be overcome. Because of the huge quantities of data involved, generating images of 3D voxel data is a very compute-intensive task, and much current work is concerned with speeding it up, using parallel processing techniques and improved algorithms. Of course, the promise of ever faster processors offers only a partial solution: as soon as you obtain a faster CPU, yet more data will arrive.
Another exciting challenge is to create a single coherent database, perhaps distributed worldwide over the Web, where data on the geometry of anatomical features are corellated with actual 2D and 3D images, and other information describing their function and properties. Perhaps soon, neuro-surgeons will wear lightweight 'Virtual Reality' spectacles which superimpose computer-generated brain images and simulations over their field of view, perfectly optically registered to the actual brain of the patient on the operating table.
Today, computers play a vital role in the understanding of brain function, and in particular, computer graphics gives us the ability to visualize what 'we' are in ways which have never before been possible. This is an exciting science without predictable end; as we delve deeper into the mysteries of our own brains there is no telling what we shall uncover. The likelihood remains, however, that whatever we find will be incomprehensible to us without the increasing power of computer visualisation.
An excellect collection of links about neuroimaging research may be found here.
Toby Howard is a lecturer at the University of Manchester.