Correlation of free‐response and receiver‐operating‐characteristic area‐under‐the‐curve estimates: Results from independently conducted FROC/ROC studies in mammography

摘要: Purpose: From independently conducted free-response receiver operating characteristic (FROC) and (ROC) experiments, to study fixed-reader associations between three estimators: the area under alternative FROC (AFROC) curve computed from data, ROC highest rating confidence-of-disease ratings. Methods: Two hundred mammograms, 100 of which were abnormal, processed by two image-processing algorithms interpreted four radiologists paradigm. inferred-ROC data derived, using assumption. Eighteen months afterwards, images same conventional paradigm; conventional-ROC (in contrast data) obtained. (inferred, conventional) analyzed nonparametric area-under-the-curve (AUC), (AFROC curve, respectively). Pearson correlation was used quantify degree association modality-specific AUC indices standard errors bootstrap-after-bootstrap method. The magnitude correlations assessed comparison with Obuchowski-Rockette fixed reader correlations. Results: Average (with 95% confidence intervals in square brackets) were: Corr(FROC, inferred ROC) = 0.76[0.64, 0.84] > Corr(inferred ROC, =more » 0.40[0.18, 0.58] Corr (FROC, 0.32[0.16, 0.46]. Conclusions: Correlation estimates high. inferred- similar modalities for a single one estimation method, suggesting that assumption might be questionable.« less