|
Optical microscopy with microtome technique has long been used to obtain the histological information. However,it is time- consuming to calculate such parameters as areas and contours for tissues. Reconstruction of 3D structure, from thin slices of a tissue, is even harder. The main purpose of this study is to separate major regions in a rat's brain, such as hippocampus, basal ganglia and thalamus on a micrograph by the image segmentation method, calculate their areas and achieve the 3D representation of the structure. Here we designed a procedure of image segmentation processing specially for these micrographs. Our approach is to process images locally based on the morphology,assisted with noise filtering. The result was compared with those obtained from other methods, such as Sobel, Compass, Laplacian and Marr-Hildreth methods. Our new pro- cedure was evaluated by phantom images with known parameters. We also tested its applicability to medical images acquired by other instrumentation. This study also includes area calculation of separated regions by use of a series of standard references. The results indicate that our procedure produced a small error (<3%) due to edge detection in region separation for the phantom images. For micrographs of a rat's brain, our method was proved superior to other traditional methods.The micrographs of sections stained with acetylcholine esterase showed a better result than those of samples unstained or stained with other chemicals. It takes about twenty minutes for this new procedure to process a whole image. The time is prolonged due to the thresholding pro- cess. Area estimation is also quite reliable since the error of reference areas estimated by the phantom square image is less than 4%. In conclusion, this new procedure can be applied to the segmen- tation process of the micrographs of the rat's brain, and thereafter offers information about important parameters before the 3D iamge can be reconstructed.
|