Automatic recognition of nervous structures by image analysis. A pattern recognition method applied to the study of...

An automatic image pattern analysis method based on differential densitometry is presented. It is applied to histological sections of mouse cerebellum. This method allows an automatic recognition of any cerebellar microscopic field. After recognition, histological images are decomposed into homogeneous parts with respect to cerebellar cellular layers, and both densitometric patterns and partial areas of each of these layers are estimated. This method is accurate to within 94% for a learning sample and 75% for a test sample. This score can be improved by adjunct algorithms presently under study.