Table 2.
Predictors for model building. Eight radiomic features and five clinical features were selected as predictors for model building.
| Radiomic Features | Good Outcome (n = 83) |
Poor Outcome (n = 86) |
b Value | p Value |
|---|---|---|---|---|
| LLL_LBP_Uniformity | 0.18 ± 0.05 | 0.20 ± 0.05 | −4.57 | 0.03 |
| LLH_Short Run Emphasis | 0.86 ± 0.08 | 0.88 ± 0.08 | −2.23 | 0.04 |
| LHL_Homogeneity 1 | 0.46 ± 0.11 | 0.40 ± 0.12 | 2.02 | 0.02 |
| HLL_Homogeneity 1 | 0.41 ± 0.09 | 0.37 ± 0.09 | 2.58 | 0.01 |
| HLL_Short Run Emphasis | 0.90 ± 0.05 | 0.92 ± 0.04 | −4.22 | 0.04 |
| HLH_Inverse variance | 0.34 ± 0.07 | 0.29 ± 0.09 | 3.45 | 0.01 |
| HLH_Short Run Emphasis | 0.90 ± 0.04 | 0.92 ± 0.04 | −4.77 | 0.02 |
| HHH_Correlation | 0.04 ± 0.06 | 0.05 ± 0.06 | 3.07 | 0.05 |
| Clinical Features |
Good Outcome
(n = 83) |
Poor Outcome
(n = 86) |
chi2 Value | p Value |
| Histology | 1 [1–2] | 1 [1–2] | 8.22 | 0.08 |
| Clinical T stage | 3 [2–4] | 3 [2–4] | 50.47 | <0.001 |
| Clinical N stage | 2 [2–3] | 1 [0–2] | 15.74 | 0.03 |
| Clinical stage | IV [III–IV] | IV [II–IV] | 17.17 | 0.02 |
| Surgery | 0 [0–0] | 0 [0–1] | 7.18 | 0.07 |
Cox regression was applied on radiomic features, and clinical features were assessed using the chi-square test. The b coefficients were assessed by Cox regression. Negative coefficients indicated decreased hazard and increased survival times. L represents a low-pass filter, and H represents a high-pass filter of the wavelet decomposition. The combination of L and H letters stands for the filter type applied to the three image axes in order. LBP, local binary pattern.