Image processing algorithms for the automated micronucleus assay in binucleated human lymphocytes

Abstract
The frequency of micronuclei in binucleated lymphocytes (cytochalasin B assay) may serve as a biological dosimeter after radiation exposure. The automation of the micronucleus assay in binucleated human lymphocytes has been considerably advanced in recent years. In our studies for this purpose the detection of binucleated cells (BNCs) and the scoring of micronuclei (MN) was divided into two parts. First, detection of BNCs was feasible with low microscopic magnification (× 100). The positions of classified BNCs were stored. Second, after an automatic change of microscope objective, the stored BNCs were automatically analyzed in sequence at high microscopic magnification (× 630) for occurrence of MN. For both phases of image analysis we used empirical methods based on mathematical morphology. The system is able to recognize nearly 65% of BNCs with false positive decisions of 6% and about 75% of the MN with false positive decisions of 7%.