Table 8 Impact of data fusion on model accuracy.
From: Secure edge-based IoMT framework for ICU monitoring with TinyML and post-quantum cryptography
Feature set | Accuracy (%) |
|---|---|
Raw features only (SpO2, HR, Temp, BP, RR) | 95.3 |
+ Rolling trends (SpO2 Trend, HR Trend, etc.) | 97.7 |
+ Physiological ratios (SpO2-to-BP, Temp-to-SpO2, etc.) | 99.4 |
