TABLE 6.
Comparison of the proposed ML-based Zao and CT estimators to prior art.
| Method | Estimated Zao (mmHg.s/mL) | Reference Zao (mmHg.s/mL) | Error (%) | r | Bland-altman biases (mmHg.s/mL) |
| Time-derivative peaks method* | 0.044 ± 0.008 | 0.056 ± 0.014 | Min: –55 Max: –20 | 0.66 | –0.011 [–0.030, 0.007] |
| Peak flow method* | 0.071 ± 0.018 | Min: –14 Max: 107 | 0.79 | 0.016 [–0.007, 0.038] | |
| PWV-based RFR | 0.056 ± 0.013 | Min: –27, Max: 34 | 0.89 | –0.000 [–0.012, 0.012] | |
| PWV-based ANN | 0.056 ± 0.012 | Min: –21, Max: 37 | 0.90 | –0.000 [–0.013, 0.010] | |
|
| |||||
| Estimated CT (mL/mmHg) | Errors (%) | Bland-altman biases (mL/mmHg) | |||
|
| |||||
| Decay time method* | 1.60 ± 0.71 | 1.16 ± 0.51 | Min: –10, Max: 152 | 0.93 | 0.44 [–0.16, 1.05] |
| Pulse pressure method* | 1.23 ± 0.48 | Min: –24, Max: 58 | 0.94 | 0.07 [–0.27, 0.41] | |
| PWV-based RFR | 1.16 ± 0.45 | Min: –36, Max: 62 | 0.93 | –0.01 [–0.39, 0.37] | |
| PWV-based ANN | 1.18 ± 0.48 | Min: –28, Max: 57 | 0.95 | 0.01 [–0.32, 0.33] | |
Zao, aortic characteristic impedance; CT, total arterial compliance.
Errors are expressed as 100 x (Xest–Xactual)/Xactual, where X is the target output under test (i.e., Zao or CT). Limits of agreement are defined as [mean bias + 1.96 SD, mean bias–1.96 SD].
*These methods utilize both the central blood pressure and flow waves for estimating Zao and CT.