Table 1 Performance comparison of the proposed method and human clinicians on the test dataset. AUROC is the area under receiver operating characteristic curve, F1 score is the harmonic mean of the precision and recall, PPV is the positive predictive value, and NPV is the negative predictive value. The performance of DeNTNet was measured with various operating point settings.

From: DeNTNet: Deep Neural Transfer Network for the detection of periodontal bone loss using panoramic dental radiographs

Performance MeasureAUROCF1 scoreSensitivitySpecificityPPVNPV
Clinician 10.840.690.740.930.650.95
Clinician 20.840.680.750.920.610.96
Clinician 30.850.680.800.910.590.96
Clinician 40.870.700.830.910.610.97
Clinician 50.850.700.780.920.640.96
Clinician Average0.850.690.780.920.620.96
DeNTNet(Balanced setting)0.950.750.770.950.730.96
DeNTNet(High sensitivity setting)0.950.710.870.900.600.97
DeNTNet(High specificity setting)0.950.730.740.960.770.95