Table 2 Summary of prognostic results from the cross-validated Kaplan Meier estimation with different TIL subtypes and feature types on the QIF datasets D3 and D4.

From: Image analysis reveals molecularly distinct patterns of TILs in NSCLC associated with treatment outcome

Dataset

Feature type

TIL subtype (s)

Cross-validated results

LUAD (D3) N = 62

Density measures

CD4

HR = 2.22 (1.11–4.45), p = 0.024, permutation significance (PS) = 0.02, C-index = 0.638 (95% CI 0.524–0.752, SE = 0.058)

CD8

HR = 2.16 (1.08–4.31), p = 0.035, PS = 0.03, C-index = 0.618 (95% CI 0.442–0.794, SE = 0.09)

LUSC (D4) N = 21

Spatial interaction

CD4 versus CD8

HR = 4.09 (1.23–13.63), p = 0.01, PS = 0.14, C-index = 0.646 (95% CI 0.464–0.828, SE = 0.093)

CD4 versus CD20

HR = 3.24 (1.01–10.40), p = 0.038, PS = 0.06, C-index = 0.637 (95% CI 0.433–0.841, SE = 0.104)

Graph- and shape-based features

CD8

HR = 3.80 (1.16–12.50), p = 0.017, PS = 0.09, C-index = 0.561 (95% CI 0.42–0.702, SE = 0.072)

CD4 plus CD20

HR = 3.43 (1.06–11.10), p = 0.029, PS = 0.10, C-index = 0.567 (95% CI 0.351–0.783, SE = 0.11)

Back to article page