Visual Processing in the Human Cerebral Cortex:
Neuromagnetic Studies



Imagine a robot looking out through a window on a sunny September day. The lights and shadows are constantly changing due to wind hitting the nearby trees. The leaves on these trees, both green and yellow, move forcefully, partially occluding the view to more distant objects and people. The walls of a close building, covered with barely detectable red bricks, reflect light steadily on the side of the shadow while around the corner the luminance fluctuates in pace of the slowly drifting clouds. After a few fast eye movements, the robot is able to construct a model of the environment by comparing the incoming sensory information to its earlier experiences. It is not only able to report which light and which shadowy areas are part of the same line or surface, but also which surfaces physically belong together being parts of the same object. This is true even for the leaves on the trees, some of which have been moving through the period of data collection. In addition, the abstract model comprises relations between the objects; "those leaves belong to the birch on the left", or, "the distant reflections come from a window of a building". Furthermore, the robot can selectively find objects and information which are relevant for its current needs, and eventually, accomplish goal- directed behavior.

We know that engineers cannot build this kind of robots - yet. This robot, the brain, is a result of natural selection, developed with ultimate pragmatism. A11 its structures and functions are built or guided by proteins, the bricks of nature capable of huge variety of configurations. One of the key structures are the cells, each of which comprises a complex chemical factory and is capable of several logical operations. The cells dispatch small and large molecules to send specific messages to neighbouring cells, and they utilize electric potentials and currents for transfer of information over longer distances along their cellular membranes. The basic functions of the brain are limited by its phylogenetic history, although all possibilities coded in the genetic prograrn may be used in parallel to achieve each task. This organ has absolutely no reason not to use all functions available to reach its goals.

Now a human-a homunculus inside one type of brain-asks: how does this machine work? Answers to this question may tell us about the relation between the robot computing a difficult scenery and a human being experiencing the results of the computation with no effort. A vast number of people have given enormous endeavor to better understand the brain processes at different levels of explanation. This study is based on one of the several available noninvasive windows into the brain. Our tool, magnetoencephalography, is able to detect electric currents within nerve cells by measuring the weak magnetic fields associated with these currents. It is able to show when, how much, and in which direction, the current is flowing. In addition, the locations of the source currents may be estimated from the measured fields. As much as we gain from the use of different senses, brain research benefits from the use of different research tools. Thus, the results of this thesis are compared mainly with results achieved with other methods.

The study of visual processing was a conscious choice. Vision is the best known sensory system in primates and enables the experimenter to present complex information fast. The extensive parcellation of subprocesses within the cerebral cortex gives a good opportunity to study cooperation between distinct areas. I hope that this study is one step in the long journey towards better understanding of who is watching through the window.