Table 3.

CRM estimation results for cross-validation training and testing for the GB-tree model on the full data set (includes the baseline, depressurization, repressurization, and recovery phases from each experiment as shown in Figure 1), followed by training and testing on only the depressurization and repressurization phases.

Full Procedure (Baseline + Depressurization + Repressurization + Recovery)
Feature Set	Normalization	Performance Metrics a
		RMSE (%)	R 2
All Features	None	13 ± 2	0.85 ± 0.04
	Baseline	12 ± 3	0.85 ± 0.08
Vital Signs	None	23 ± 3	0.50 ± 0.14
	Baseline	17 ± 2	0.71 ± 0.08
ABP Waveform	None	13 ± 2	0.85 ± 0.04
	Baseline	13 ± 3	0.84 ± 0.08
HRDN only	None	14 ± 3	0.82 ± 0.07
	Baseline	15 ± 5	0.77 ± 0.04
Simulated Hemorrhage (Depressurization)
Feature Set	Normalization	Performance Metrics a
		RMSE (%)	R 2
All Features	None	14 ± 2	0.74 ± 0.11
	Baseline	13 ± 4	0.77 ± 0.09
Vital Signs	None	18 ± 3	0.56 ± 0.13
	Baseline	17 ± 2	0.65 ± 0.07
ABP Waveform	None	14 ± 2	0.76 ± 0.10
	Baseline	13 ± 3	0.78 ± 0.09
HRDN only	None	16 ± 3	0.67 ± 0.12
	Baseline	16 ± 3	0.68 ± 0.13
Simulated Resuscitation (Repressurization)
Feature Set	Normalization	Performance Metrics a
		RMSE (%)	R 2
All Features	None	14 ± 3	0.74 ± 0.11
	Baseline	16 ± 4	0.67 ± 0.19
Vital Signs	None	21 ± 3	0.44 ± 0.16
	Baseline	15 ± 2	0.71 ± 0.08
ABP Waveform	None	14 ± 3	0.75 ± 0.08
	Baseline	17 ± 6	0.71 ± 0.08
HRDN only	None	13 ± 3	0.80 ± 0.09
	Baseline	18 ± 8	0.56 ± 0.37

^a. Metrics presented as test set mean ± standard deviation after 5-fold cross validation.