Table 1 Blinded pathologist reader study of diagnostic concordance

From: Deep learning-enabled realistic virtual histology with ultraviolet photoacoustic remote sensing microscopy

Breast tissue (n = 24 pairs)

Malignancy prevalence = 0.42

Ground Truth Fleiss’ Kappa* = 1.00

 

P1

P2

P3

P4

P5

Mean

Consensus

Sensitivity

0.90

0.90

1.00

1.00

1.00

0.96

1.00

Specificity

0.93

0.86

1.00

0.86

0.93

0.91

0.93

Positive predictive value

0.90

0.82

1.00

0.83

0.91

0.89

0.91

Negative predictive value

0.93

0.92

1.00

1.00

1.00

0.97

1.00

Accuracy

0.92

0.88

1.00

0.92

0.96

0.93

0.96

Concordance (κ)†

0.83

0.75

1.00

0.83

0.92

0.86

–

Prostate tissue (n = 32 pairs)

Malignancy prevalence = 0.63

Ground Truth Fleiss’ Kappa* = 0.81

 

P1

P2

P3

Mean

Consensus

Sensitivity

0.80

0.95

0.85

0.87

0.85

Specificity

1.00

0.83

1.00

0.94

1.00

Positive Predictive Value

1.00

0.90

1.00

0.97

1.00

Negative Predictive Value

0.75

0.91

0.80

0.82

0.80

Accuracy

0.88

0.91

0.91

0.90

0.91

Concordance (κ)†

0.75

0.80

0.81

0.79

–

  1. Summary statistics from diagnostic concordance studies for breast and prostate tissue where a panel of pathologists (P) were tasked with interpreting paired virtual histology and ground truth H&E-stained brightfield images.
  2. †Cohen’s kappa for intra-observer concordance.
  3. *Fleiss’ kappa measures inter-observer concordance for interpreting ground truth H&E images. Source data are provided as a Source Data file.
Back to article page