Table 3.
Summary of modeling and application in ICP signal.
| References | Application | Features or information | Modeling approach | Performance or delivery |
|---|---|---|---|---|
| Quachtran et al. (81) | IHD | CNN based features | Regression; CNN, CNN + Autoencoder * | ACC: 92.05% |
| Scalzo et al. (82), Scalzo and Hu (39) | IHD | MOCAIP Metrics improved by CDF; Trending features | Threshold-based; SR-DA, SVM, SR-KDA* | AUC: 85.9%; Reduce FPR by 31% |
| Soehle et al. (71) | IHD | Complexity (scaling exponent, SampEn, multiscale entropy) | Statistical analysis | scaling exponent (p < 0.001), SampEn (p = 0.004), MSE (p < 0.05) |
| Scalzo et al. (41) | IHP | MOCAIP Metrics with backward sequential feature selection | MLR; Adaboost; Extra-Tree*; | AUC: 96%, SPE: 98%, SEN: 93% (1 min prior to the elevation) |
| Xiao et al. (80) | IHP | Optimal MOCAIP metrics found by DE algorithms | Regularized quadratic discriminator | SPE: 99%, SEN: 37% (5 min prior to the elevation) |
| Hamilton et al. (83) | IHP | Top 10 MOCAIP metrics found by PSO | A quadratic classifier (QDC) | ACC: 77%, SEN: 90%, SPE: 75% (5 min prior to the elevation) |
| Hornero et al. (73) | IHD | ApEn | Statistical analysis: non-parametric bootstrap hypothesis testing | p < 0.01 (IH group vs. recovering group) |
| Teplan et al. (50) | LOP | 7 Time domain features | hGMMs, ROC analysis | YI = 0.33, SPE: 85%, SEN: 48% |
| Pimentel et al. (84) | LOP | GPs based dynamic features, PRx-based statistical features | GPs-based, PRx-based, Combined Model* | ACC: 74%, SPE: 65%, SEN: 83%, AUC: 76% |
| Güiza et al. (54) | LOE | Insult Intensity (I), Insult duration (D), LAx | Multivariate logistic regression models and color-coded plot | ICP- time burden Visualization, |
| Howells et al. (8) | LOE | Pressure reactivity indices in various frequency band | Statistical analysis: correlation, Spearman's R, two-tailed Wilcoxon matched pairs test | ρ = −0.46 (correlation with GOSe) |
| Lazaridis et al. (85) | LOE | PRx, Cumulative area under the curve above threshold | Logistic regression models, | AUC: 0.77, 95% CI 0.70–0.83 |
| Lu et al. (75) | LOE and LOP | Multiscale entropy | Statistical analysis: ANOVA, Logistic regression | Favorable (F = 28.7, p < 0.0001), Unfavorable (F = 17.21, p < 0.0001) ACC: 82%, SPE: 94%, SEN: 50% |
IHD, Intracranial Hypertension Detection; IHP, Intracranial Hypertension Prediction; LOE, Long-term Outcome Evaluation; LOP, Long-term Outcome Prediction; ACC, Accuracy; SPE, Specificity; SEN, Sensitivity; AUC, Area Under receiver operating characteristic Curve; CNN, Convolutional Neural Networks; DE, differential evolution; hGMMs, hierarchical Gaussian mixture models; YI, Youden index; GPs, Gaussian processes; Lax, low-frequency autoregulation index; GOSe, extended Glasgow Out- come Scale; ANOVA, Analysis of Variance; ApEn, Approximate Entropy; PSO, Particle Swarm Optimization; SR-DA, Spectral regression- discriminant analysis; SR-KDA, Spectral Regression-Kernel Spectral Regression.
is the algorithm with best performance.