Radiomics and Machine Learning Differentiate Soft-Tissue Lipoma and Liposarcoma Better than Musculoskeletal Radiologists

Table 3

Accuracies of radiologists and radiomics model in identifying liposarcoma versus lipoma.

	RMX (%)	RAD 1 (%)	RAD 2 (%)	RAD 3 (%)	Consensus (%)
Accuracy	94.7	65.8	81.6	79.0	81.6
Sensitivity	88.8	76.9	76.9	76.9	76.9
Specificity	100	60.0	84.0	80.0	84.0
PPV	100	50.0	71.4	66.7	71.4
NPV	78.5	83.3	87.5	87.0	87.5
AUC	0.926	0.685	0.805	0.785	0.805

RMX: radiomics model; RAD 1, 2, and 3: MSK radiologists 1, 2, and 3; consensus: group consensus between the three MSK radiologists; PPV: positive predicting value; NPV: negative predicting value; AUC: area under the receiver operating curve.