| Abstract: | A quantitative image analysis of bone marrow blast cells from eight patients with refractory anemia with an excess of blast cells was performed using the SAMBA 200 cell image analyzer. A total of 33 parameters was computed on 665 cells visually classified as B1 blasts (agranular cells), B2 blasts (cells containing a few azurophilic granules) and B3 blasts (cells with numerous granules). The continuum of variation of some cytoplasmic parameters (area, hue and standard deviation of the luminance, hue and saturation histograms) and some nuclear parameters (area and convexity degree) from the B1 cells to the B3 cells indicated a concomitant cytoplasmic and nuclear maturation, with the B1 cells being the most immature. An attempt to automatically classify these cells using a stepwise linear discriminant analysis resulted in an average of 68% correctly classified cells at the fifth step. Among the B1 and B2 blast cells considered together, an unsupervised classification method distinguished seven subgroups of blasts, which were principally different in cytoplasmic area, cytoplasmic color and nuclear texture. The percentages of cells belonging to two of these subgroups were highly discriminatory with respect to the prognosis. These two cell types had morphologic, textural and color features that put them very near the normal immature myeloidlike progenitor cells and normal myeloblasts, as demonstrated by means of canonical analysis. All patients having a very low percentage of these two cells among their blast cells died from overt leukemia less than one year after the first diagnosis of their disease. |
