Article Figures & Data
Figures
- Fig 1.
The schematic diagram for data processing and analysis.
- Fig 2.
ROC curves of the different feature combinations with the SVM classifier (CM, RLM, and Hist stand for the co-occurrence matrix, run-length matrix, and histogram-based texture types, respectively).
- Fig 3.
Optimal feature subset selection. A, Selection process of the optimal feature subset with RFE-SVM. B, ROC curve of the optimal feature subset selected.
- Fig 4.
Ranking orders of all features in the optimal subset. The length of each bar represents the iteration number at which the feature was rejected by the RFE.
- Fig 5.
Kaplan-Meier survival curves determined by the log-rank test (P < .001).
Tables
Group Sex (Male/Female) Age (Median, Range) (yr) Overall Survival (Median, Range) (days) Long-term group 40:27 59, 14∼81 600, 370∼1731 Short-term group 43:23 64, 17∼85 145, 16∼357 Feature Group/Feature ID Description CM features CM1 Energy CM2 Contrast CM3 Correlation CM4 Variance CM5 Inverse difference moment CM6 Sum average CM7 Sum variance CM8 Sum entropy CM9 Entropy CM10 Difference variance CM11 Difference entropy CM12 Information measures I of correlation CM13 Information measures II of correlation RCM1–RCM13 Range of corresponding features listed above RLM features RLM1 Short-run emphasis RLM2 Long-run emphasis RLM3 Gray-level nonuniformity RLM4 Run-length nonuniformity RLM5 Run percentage RLM6 Low gray-level run emphasis RLM7 High gray-level run emphasis RLM8 Short-run low gray-level emphasis RLM9 Short-run high gray-level emphasis RLM10 Long-run low gray-level emphasis RLM11 Long-run high gray-level emphasis RRLM1–RRLM11 Range of corresponding features listed above Histogram-based features H1 Mean H2 Entropy H3 Uniformity H4 SD H5 Smoothness H6 Skewness H7 Third-order moment H8 Kurtosis Note:—ID indicates identification; RCM, range of co-occurrence matrix; RRLM, range of run-length matrix; H, histogram.
- Table 3:
The performance of feature combinations with the SVM classifier in differentiating long- and short-term survival time
Feature Set NFa SVM Sensitivity Specificity Accuracy AUC CMb 26 0.7910 0.6667 0.7293 0.7786 RLM 22 0.7313 0.4242 0.5789 0.5880 Histogram 8 0.6866 0.2576 0.4737 0.4326 CM + RLMb 48 0.7761 0.7727 0.7744 0.8053 CM + histogram 34 0.7761 0.6818 0.7293 0.7782 RLM + histogram 30 0.8060 0.3788 0.5940 0.6223 CM + RLM + histogramc 56 0.7761 0.7879 0.7820 0.8057
