Computer Techniques in the Analysis of Electrophysiological Data from the Human Thalamus

Abstract

Stereotaxic surgery poses problems in three-dimensional geometry which can be solved on-line by a computer during an operation. Data processing techniques can be applied to the storage and recall of anatomical data during an operation. Stimulation and recording from the thalamus and neighbouring structures during surgery give a large variety of data which require computer processing. Stimulation gives subjective sensory phenomena which vary in quality and intensity and are referred to different parts of the body, and also to motor phenomena; micro-electrode recording gives different responses in neurons to different kinds of peripheral stimuli. Each report of subjective, motor and neuronal behaviour is an item of data located in a space defined by a co-ordinate system. There are several different co-ordinate systems that can be used, varying in the choice of reference planes and zero points and in the use of absolute or proportionate measure. Thus the variability in data has contributions, not only from many different aspects of technique and reporting, it also represents individual differences, possibly in part hereditary between one patient and another, and differences in their attentive and linguistic habits and in their response to structural changes in disease. The consequence of this great variety both in the phenomena themselves and in their relative and absolute localization is that the set of data on one patient at one operation never exactly corresponds to any other set and the data from two different surgical departments are not easy to compare. But without more comparison and pooling of data no generalisations are possible. The central problem in making this volume of non-repeatable data comparable is to make each item in a form suitable for computer processing. When this is done, the systematic features can be separated from the random ones and a three-dimensional map of the functional or physiological properties of the thalamus and neighbouring structures can be constructed. When such a map has been made it can be superimposed on an independent mapping of structural features on the same co-ordinate system. Both the structural and functional plottings will be of probabilities, and the final requirements of a scientific analysis will only be met when hypotheses can be shown to survive properly constructed statistical tests of significance. The steps to achieve this can be divided into various stages. The first involves the preparation of data from the operating room written on forms as numbers and text into a format to make it acceptable as computer input. Those parts in text present the first problem. Decisions have to be made on reducing the vocabulary and it has been found that about 100 words of vocabulary together with a simplified syntax give a sufficiently flexible coding for most items without significant loss of subtlety. However, provision has to be made for entry into the computer as original text so that this can be recalled in exceptional cases. Examples will be given of solutions to this problem of text preparation on data from stimulation and recording. The next step is to make all the coded items of data compatible in respect to their three-dimensional localization and when this has been done data from different centres can be pooled. The computer is ideally suited to this task because it can be programmed to carry out the operations of transposition of axes and zero points and changes of scale from any defined co-ordinate system to any other system. The third stage is to devise significance tests to examine for systematic features in the data. At this point a difficulty is encountered as the stereotactic operation involves an approach to certain defined target points. This means that the sampling of the deep structures is not spatially random and the degree of biasing has to be discovered from the data. The next stage is to prepare three-dimensional probability plots and for these the ''negative'' information in the data are needed that is to say the number of times a particular region has been explored with negative findings. The computer is well suited to recover this information. The final stage is the super-imposition of these functional probability maps with the probabilities based on anatomical data. Examples of this work and the pooling of data from three independent centres will be discussed.