The characterization of biological features in dedifferentiated liposarcomas by means of principal components and discriminant analyses of 25 computer-generated variables from Feulgen-stained nuclei and histological slides.

Several groups of lipomatous tumors are not yet clearly characterized on the biological level. In order to attempt to classify the dedifferentiated liposarcomas with respect to other types of malignant liposarcomas, 80 adipose tumors were submitted to the combination of two computer-assisted methodologies. These two methodologies consisted of i) the determination of 25 variables, and ii) the analysis of the diagnostic information contributed by these 25 variables by means of two complementary techniques, i.e. principal components and discriminant analyses. The 25 variables were computed by means of image cytometry on Feulgen-stained nuclei and histological slides, quantitatively describing distinct biological characteristics relating to morphonuclear (chromatin pattern) features (14 variables), nuclear DNA content distribution (9 variables), and tissue architecture pattern (2 variables). The 80 adipose tumors included 21 typical lipomas, 7 atypical lipomas (defined as extremity adipose tumors with a histopathological pattern of well-differentiated liposarcomas), 16 retroperitoneal and 5 non-retroperitoneal abdominal well-differentiated liposarcomas, 9 dedifferentiated liposarcomas, 8 myxoid (intermediate-grade tumor) and 14 pleomorphic (high-grade tumor) liposarcomas. The data strongly suggest that the dedifferentiated liposarcomas exhibit biological characteristics which are distinct from those of low- and high-grade liposarcomas, but similar to those of intermediate ones. The results also show that typical and atypical lipomas are two distinct biological entities. In contrast, the atypical lipomas and the well-differentiated retroperitoneal and non-retroperitoneal liposarcomas exhibited a high number of similar biological characteristics. Computer-assisted methods contribute valuable information to characterize lipomatous tumor biology.