Diagnostic performance of ADC maps capable of distinguishing HPV+ from HPV– OPSCC on the basis of ADC mean, skewness, and kurtosisa
Feature | AUC | P Value | TP | FP | TN | FN | Sensitivity | Specificity | Accuracy | Optimal Threshold |
---|---|---|---|---|---|---|---|---|---|---|
ADCb0–1000 | ||||||||||
ADC mean | 0.723 | .019 | 8 | 7 | 16 | 3 | 0.727 | 0.695 | 0.706 | 1062.429b |
ADC skewness | 0.731 | .016 | 8 | 6 | 17 | 3 | 0.727 | 0.739 | 0.735 | 0.349 |
ADC kurtosisc | 0.893 | <.001 | 11 | 4 | 19 | 0 | 1.000 | 0.826 | 0.882 | 0.640 |
ADCb0–750–1000 | ||||||||||
ADC mean | 0.755 | .009 | 8 | 7 | 16 | 3 | 0.727 | 0.695 | 0.706 | 1088.702b |
ADC skewness | 0.715 | .023 | 8 | 5 | 18 | 3 | 0.727 | 0.783 | 0.765 | 0.500 |
ADC kurtosisc | 0.826 | .001 | 10 | 5 | 18 | 1 | 0.909 | 0.783 | 0.823 | 0.651 |
ADCb0–500–750–1000 | ||||||||||
ADC mean | 0.755 | .009 | 8 | 5 | 18 | 3 | 0.727 | 0.783 | 0.765 | 1034.222b |
ADC skewness | 0.727 | .009 | 8 | 4 | 19 | 3 | 0.727 | 0.826 | 0.794 | 0.525 |
ADC kurtosisc | 0.826 | .001 | 10 | 5 | 18 | 1 | 0.909 | 0.783 | 0.823 | 0.620 |
ADCb100–1000 | ||||||||||
ADC mean | 0.636 | .105 | 8 | 8 | 15 | 3 | 0.727 | 0.652 | 0.676 | 958.089b |
ADC skewness | 0.581 | .231 | 7 | 9 | 14 | 4 | 0.636 | 0.609 | 0.618 | 0.133 |
ADC kurtosisc | 0.751 | .010 | 9 | 7 | 16 | 2 | 0.818 | 0.696 | 0.735 | 0.484 |
Note:—TP indicates true-positive; FP, false-positive; TN, true-negative; FN, false-negative.
↵a For comparison, the diagnostic performance of the perfusion-insensitive map recommended in the literature14 is equally shown. P values to distinguish HPV+ from HPV– OPSCC were calculated with the Mann-Whitney-Wilcoxon test. Sensitivity, specificity, and accuracy were calculated using the optimal threshold (Youden index from receiver operating curve analysis).
↵b ADC mean thresholds in x10−6mm2/s.
↵c Excess kurtosis.