Developing a diagnostic model to differentiate the well-differentiated lipomatous tumors based on clinicopathological characteristics

Fig. 1. Microscopic Characteristics of Well-Differentiated Lipomatous Tumors. Lipoma: (A) Composed of well-differentiated adipocytes with tightly arranged tumor cells. (B) Tumor lobulated by fibrous septa into variably sized adipose lobules. (C) No MDM2 gene amplification observed. Atypical Lipomatous Tumors/Well-Differentiated Liposarcomas (ALT/WDLPS): (D) Adipocytes appear round or polygonal, with large lipid droplets displacing the cytoplasm and nuclei toward the periphery. Nuclei are flattened and oval. (E) Mature adipocytes were observed without mitotic activity or necrosis. (F) Cluster-type MDM2 gene amplification present.

Fig. 2. ROC Curve Analysis of Tumor Maximum Diameter and MDM2 Gene Amplification in Well-Differentiated Lipomatous Tumors. The ROC curve analysis demonstrated that a maximum tumor diameter of 9.9 cm served as the optimal cut-off value for predicting MDM2 amplification, yielding a sensitivity of 0.825, specificity of 0.778, Youden index of 0.603, and an area under the curve (AUC) of 0.855.

Fig. 3. Construction of the Integrated Diagnostic Model for Atypical Lipomatous Tumors/Well-Differentiated Liposarcomas (ALT/WDLPS). (A) The model incorporated eight predictive factors, each assigned a weighted score on a linear scale based on its association with malignancy risk. The total score was used to estimate the probability of ALT/WDLPS. (B–C) Risk scores derived from the model were subjected to K-means clustering in the training set (B). Two cut-off values (0.219 and 0.652) were identified from the midpoints between adjacent cluster centers, stratifying patients into low-, intermediate-, and high-risk groups. This stratification was validated in the test set (C). (D) ROC curve analysis showed excellent diagnostic performance, with a sensitivity of 0.933, specificity of 0.722, and an AUC of 0.928.