FISH AND CHIPS - AUTOMATION OF FLUORESCENT DOT COUNTING IN INTERPHASECELL-NUCLEI

Fluorescence in situ hybridization allows the enumeration of chromosom al abnormalities in interphase cell nuclei. This process is called dot counting. To estimate the distribution of chromosomes per cell, a lar ge number of cells have to be analyzed, especially when the frequency of aberrant cells is low. Automation of dot counting is required becau se manual counting is tedious, fatiguing, and time-consuming. We devel oped a completely automated fluorescence microscope system that can ex amine 500 cells in approximately 15 min to determine the number of lab eled chromosomes (seen as dots) in each cell. nucleus. This system wor ks with two fluorescent dyes, one for the DNA hybridization dots and o ne for the cell nucleus. After the stage has moved to a new field, the image is automatically focused, acquired by a Photometrics KAF 1400 c amera (Photometrics Ltd., Tuscon, AZ, USA), and then analyzed on a Mac intosh Quadra 840AV (Apple Computer, Inc., Cupertino, CA, USA) compute r. After the required number of cells has been analyzed, the user may interact to correct the computer by working with a gallery of the cell images. The automated dot counter has been tested on a number of norm al specimens where 4,'6-diamidino-2-phenylindole (DAPI) was used for t he nucleus counterstain and a centromeric 8 probe was used to mark the desired chromosome. The slides contained lymphocytes from cultured bl ood. We compared the results of the dot counter with manual counting. Manually obtained results, published in the literature, were used as t he ''ground truth.'' For a normal specimen, 97.5% of cells will have t wo dots. Fully automated scanning of 13 slides showed that an average of 89% of all nuclei were counted correctly. In other words, an averag e of 11% has to be interactively corrected using a monitor display. Th e machine accuracies, after interactive correction, are comparable to panels of human experts (manual). The fully automatically obtained res ults are biased with respect to manual counting. An error analysis is given, and different causes are discussed. (C) 1997 Wiley-Liss, Inc.