Cancer Imaging

Table 3 Model performances in the training and external validation cohorts

From: Deep learning radiomics-based prediction model of metachronous distant metastasis following curative resection for retroperitoneal leiomyosarcoma: a bicentric study

Model	Training Cohort					External Validation Cohort
Model	SPE	NPV	ACC	AUC	95% CI	SPE	NPV	ACC	AUC	95% CI
DLRN	0.959	0.868	0.884	0.939	0.905–0.974	0.814	0.921	0.810	0.822	0.692–0.953
DLRS	0.918	0.870	0.868	0.937	0.899–0.975	0.905	0.864	0.810	0.786	0.649–0.923
RS	0.904	0.864	0.860	0.917	0.870–0.964	0.780	0.889	0.741	0.733	0.573–0.892
Clinical model	0.875	0.872	0.686	0.718	0.644–0.793	0.674	0.750	0.500	0.511	0.359–0.662
DeLong test	Standard Error		95% CI		P value	Standard Error		95% CI		P value
DLRN vs. DLRS	0.009		-0.017–0.022		0.816	0.038		-0.037–0.083		0.465
DLRN vs. RS	0.109		-0.160–0.267		0.624	0.031		-0.024–0.096		0.233
DLRN vs. Clinical	0.037		0.148–0.294		< 0.001	0.092		0.094–0.456		0.003
DLRS vs. Clinical	0.042		0.137–0.301		< 0.001	0.109		0.098–0.525		0.004
DLRS vs. RS	0.114		-0.133–0.313		0.429	0.032		-0.042–0.083		0.523
RS vs. Clinical	0.045		0.110–0.287		< 0.001	0.091		0.043–0.401		0.015

Note: The AUCs among models were compared using the DeLong test
Abbreviations: DLRN, deep learning radiomic nomogram; DLRS, deep learning radiomics signature; RS, radiomics signature; SPE, specificity; NPV negative predictive value; ACC, accuracy; AUC, area under the receiver operating characteristic curve; CI, confidence interval

Back to article page

ISSN: 1470-7330

Contact us

Submission enquiries: alyzza.javate@springernature.com
General enquiries: info@biomedcentral.com