Table 6.
The pairwise classification results
| Groups | Feature set | Accuracy | Sensitivity | Specificity | |
|---|---|---|---|---|---|
| SIRS vs Sepsis | sign_m | WE(s2) | 40% | 80% | 20% |
| sign_m | WE(s2), WEn(s6) | 80% | 80% | 80% | |
| sign_mdetr | CWTentro4 | 80% | 80% | 80% | |
| sign_mdetr | CWTentro4, WE(s5), CWTene | 93.33% | 100% | 90% | |
| SIRS vs S. Shock | sign_m | WEn(s5) | 83.33% | 80% | 85.71% |
| sign_m | WEn(s5), WEn(s6), WEn(s8), WE(s8) | 91.67% | 80% | 100% | |
| sign_mdetr | WEn(s6) | 83.33% | 80% | 85.71% | |
| sign_mdetr | WEn(s6), CWTentro4, WE(s8) | 100% | 100% | 100% | |
The table demonstrates the randset feature sets and the classification performance achieved with a linear classifier and leave-one-out cross-validation, in terms of accuracy, sensitivity, and specificity. Here, sensitivity refers to SIRS, and specificity refers to sepsis or septic shock, respectively. The results are presented for sign_m and sign_mdetr, separately. Both univariate models for the best feature selected, and multivariate models, are depicted. WE, wavelet energy; Wen, wavelet entropy; CWT, continuous wavelet transformation; s, scale; entro4, entropy per scale 4.