An accurate method for automated counting of silver grains in autoradiographs

A grain counting algorithm has been developed for use with an image analysis system interfaced with a microscope. The algorithm consists of a sequence of tests involving thresholding, prospective grain center detection, contrast, and three kinds of artifact detection. The function of each of these parameters is described. This algorithm is accurate to within 4% for those fields most frequently encountered, and to within 9% for all fields examined. Furthermore, it operates very quickly and at relatively low magnification, making it possible to scan large areas efficiently.