Skip to main content

Table 3 Performance metrics for integrated I-score (two-category) models and reference models

From: Automated whole-slide images assessment of immune infiltration in resected non-small-cell lung cancer: towards better risk-stratification

Models

Discovery cohort

 

Validation cohort

iAUCa

Harrell’s C-index

AIC

 

iAUC

Harrell’s C-index

AIC

TNM stage model b

0.674

0.694 (0.640–0.749)

615.6

 

0.645

0.651 (0.596–0.705)

742.3

I-score model

0.584

0.592 (0.532–0.651)

647.4

 

0.592

0.588 (0.533–0.644)

758.6

TNM Stage & I-score model b

0.699

0.711 (0.651–0.772)

613.5

 

0.673

0.679 (0.623–0.736)

736.4

Clinicopathologic model c

0.698

0.728 (0.676–0.781)

614.2

 

0.671

0.685 (0.627–0.743)

739.2

Full model c

0.717

0.742 (0.688–0.795)

610.9

 

0.684

0.695 (0.639–0.751)

734.8

  1. Data in parentheses are 95% confidence intervals
  2. iAUC integrated area under the curve, Harrell’s C-index Harrell’s concordance index, AIC Akaike information criterion
  3. aiAUC refers to the integrated area under the ROC curve
  4. b TNM-stage model vs. TNM Stage & I-score model: likelihood ratio P = 0.044
  5. cClinicopathologic model (TMN stage & differentiation grade & adjuvant chemotherapy) vs. Full model: likelihood ratio P = 0.022