FIGURE 6.

Receiver–operator characteristic (ROC) curves based on using sMRI QC metrics for classification. (A) ROCs for differentiating bad data from acceptable data (questionable and good). CNR best differentiated bad data from the acceptable data as it yielded the largest area under the curve (AUC = 0.92). (B) ROCs for differentiating poor (bad + questionable) data from good data. CNR and TCTV showed the best performance with AUC = 0.7–0.73. SNR, signal-to-noise ratio; SVNR, signal variance-to-noise ratio; CNR, contrast-to-noise ratio; CVNR, Contrast variance-to-noise ratio; TCTV, tissue contrast-to-tissue (intensity) variance; FWHM, full width-at-half maximum; CoM, center of mass.