Semiautomatic image analysis for grain counting in in situ hybridization experiments.

We have developed a computer image analysis procedure for counting autoradiographic grains in in situ hybridization experiments. The procedure automatically estimates the number of autoradiographic grains over cells and measures cell number and size so that grain density per unit cell area can be calculated. Advantages include the clear separation of grains and cells, using chromatic and spatial filters to enhance the image; the use of gray level operators to extract cells from grains; and the use of binary operators for separating apposed or partially overlapping cells and grains. Comparison of manual and automated grain counts revealed a significant correlation between human and computer estimations of grain number. However, the automatic grain counting technique consistently underestimated the number of grains when grain density was high. Measures of the fractional area occupied by grains normalized by the average area of a single grain were a better estimate at high grain densities. The procedure can be modified easily to operate on most image analyzers.