Tumor-to-bone distance and radiomic features on MRI distinguish intramuscular lipomas from well-differentiated liposarcomas

Table 2 Performance of machine learning model and two experienced musculoskeletal (MSK) radiologists in distinguishing IM lipomas from atypical lipomatous tumors or well-differentiated liposarcomas

	AUC (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	Accuracy (95% CI)	P value*
Machine learning model	0.88 (0.72–1)	91.6 (62.0–100)	85.7 (42.0–100)	89.0 (67.0–98.6)
MSK radiologist 1	0.94 (0.87–1)	97.4 (86.0–100)	90.9 (71.0–99.0)	95.0 (86.1–98.9)	0.4
MSK radiologist 2	0.91 (0.83–0.99)	100 (91.0–100)	81.8 (60.0–95.0)	93.3 (83.8–98.1)	0.7

AUC The area under the receiver operator characteristic curve, CI Confidence interval
*P value based on the Delong’s test, which were used to compare the AUCs between the machine learning model and radiologists

ISSN: 1749-799X