Quantitative image analysis for immunocytochemistry and in situ hybridization.

Image analysis hardware, software, and procedures are described for analysis of tissue reacted for antibody immunocytochemistry and in situ hybridization. A Magiscan image analyzer is used to process images viewed with a light microscope. LUT functions, spatial filters (parabola) and gray level convolutions (sharpen, laplacian, mexican hat) are applied in order to extract immunoreaction product or autoradiographic grains. These objects are then thresholded and binary operators (erosion, dilation, separation) are applied to separate closely apposed objects. Measurements routines are used to estimate the optical density and size of labeled profiles or to count grains and compute grain density per profile. A JEOL 1210 electron microscope is used to view tissue treated for post-embedding immunochemistry. Digital images are captured with a Kodak 1K CCD camera, archived, transported across a local area network, stored on optical disks and analyzed on a MacIntosh IIci. NIH Image is used to process these images. Results show that the optical density of GABA antibody labeling is reduced by monocular deprivation, that substance P mRNA hybridization labeling is increased by scopolamine, and that retinal terminals are densely labeled by antibodies to glutamate. These techniques are thus useful for measuring the amount of change in labeling after experimental manipulations and for distinguishing labeled from unlabeled profiles.