A preliminary study on automatic identification of histological Structures in tissue sections

Abstract

Hand-tracing is commonly used to record structural information from histological specimens. It is relatively troublesome and requires numerous steps to develop the figures and has poor precision.
In computer-aided image processing, image data is normally treated as gray-scale data, restricting the use of image processing with histological specimens. Conventionally, stained specimens like hematoxylin-eosin are difficult to use in this way.
The present study extracted structural information from hematoxylin-eosin stained specimens and treated them as full-color images. For this type of image processing, the maximum likelihood method, a static method was used in conjunction with an original program written by the authors.
As a result, all processes from the first image of the histological specimen to the final image lasted for only about 20min.
Further investigation is required to record the structure of teeth and bone automatically.