Electroencephalogram (EEG) analysis is an important means of obtaining detailed information on the brain functions which CT (computed tomography) or angiography cannot provide. To help grasp the brain functions visually and intuitively in diagnosis and treatment planning of brain diseases, the computerized topographic EEG (or topographic EEG mapping), giving a two-dimensional image of EEG amplitude or frequency distribution on the entire scalp through the use of a computer, has recently been used. At present, several types of interpolation methods are available for the. interpolation process required by the topographic EEG mapping. However, very few reports have been published regarding the evaluation of the current interpolation method.
This paper attempts to evaluate four interpolation methods from the viewpoints of interpolation accuracy and computational speed through experiments using actual EEG data. The four methods studied here are : “two-dimensional sampling function” “linear interpolation”, “bicubic lofted surface” and “Coons' bilinear curved surface”. As a result, the interpolation by Coons' bilinear curved surface is found to be most accurate among the four methods. So it is considered to be optimum when a sufficient length of time is available for computation. Accuracy of the linear interpolation method is slightly inferior compared with that of the Coons' bilinear curved surface method, but this would pose no problems in practical use. Furthermore, computational speed of the linear interpolation method is the highest among the four, then the method is considered to be most suitable when interpolation results are needed whithin a short time.