A novel algorithm for minute ventilation estimation in remote health monitoring with magnetometer plethysmography.

The purpose of this study was to evaluate the accuracy of minute ventilation (V˙E) estimation using a novel method based on a non-linear algorithm coupled with cycle-based features. The experiment protocol was well adapted for remote health monitoring applications by exploiting data streams from respiratory magnetometer plethysmography (RMP) during different physical activity (PA) types. Methods Thirteen subjects with an age distribution of 24.1±3.4 years performed thirteen PA ranging from sedentary to moderate intensity (walking at 4 and 6 km/h, running at 9 and 12 km/h, biking at 90 W and 110 W). In total, 3359 temporal segments of 10s were acquired using the Nomics RMP device while the iWorx spirometer was used for reference V˙E measurements. An artificial neural network (ANN) model based on respiration features was used to estimate V˙E and compared to the multiple linear regression (MLR) model. We also compared the subject-specific approach with the subject-independent approach. Results The ANN model using subject-specific approach achieved better accuracy for the V˙E estimation. The bias was between 0.20±0.87 and 0.78±3 l/min with the ANN model as compared to 0.73±3.19 and 4.17±2.61 l/min with the MLR model. Conclusion Our results demonstrated the pertinence of processing data streams from wearable RMP device to estimate the V˙E with sufficient accuracy for various PA types. Due to its low-complexity and real-time algorithm design, the current approach can be easily integrated into most remote health monitoring applications coupled with wearable sensors.