Quantitative analysis of disseminated tumor cells in the bone marrow by automated fluorescence image analysis

Accurate quantification of disseminated tumor cells in hematological samples is of fundamental importance in clinical oncology. However, even highly standardized protocols allow only a rough estimation of the total analyzed cell number, as sample processing may have adverse effects on the number of cells available for analysis. The fluorescence-based microscopic scanning system (MetaCyte) detects, counts, captures, and relocates immunolabeled tumor cells in hematopoietic samples. We report on a cell-counting approach that has been implemented into the scanning system to precisely quantify the number of cells per slide. The cell-counting function, which was designed to determine the number of all nucleated (DAPI-stained) cells on the slide, allows an accurate counting of the tumor cells and the total number of cells analyzed in the given microscopic sample. The reliability of the cell-counting approach was demonstrated by the analysis of DAPI-stained images with 18-1,363 nucleated cells. A good correlation (r(2) = 0.965) between the manually and automatically gained results was observed. The counting accuracy could even be optimized after implementing a correction factor. To prove or disprove an interslide variation, routine bone marrow cytospin preparations from neuroblastoma patients were immunostained for GD2/FITC and counterstained with DAPI. Automatic cell counting of cytospin preparations from the same patients showed significant differences in the total cell number (up to 67% cell loss during preparation, with a maximum interslide difference of 4.7 x 10(5) mononuclear cells). We conclude that determination of the tumor cell content in hematopoietic samples is only reliable when it is performed together with accurate cell counting.