Related Experiment Video
Updated: Jan 12, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.4K
A neural algorithm for the non-uniform and adaptive sampling of biomedical data.
1Mathematical Biology and Physiology, Dipartimento di Elettronica e Telecomunicazioni, Politecnico di Torino, Corso Duca degli Abruzzi 24, Torino, 10129 Turin, Italy.
Computers in Biology and Medicine
|February 27, 2016
Summary
This study introduces an adaptive neural network algorithm for body sensor networks that intelligently adjusts sampling rates to capture crucial health data efficiently. This method reduces data transmission and memory usage while maintaining signal accuracy, even below the Nyquist limit.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Wearable Technology
Background:
- Body sensors are increasingly used for health monitoring and remote surveillance.
- Physiological data from these sensors can be non-stationary, with bursts of high amplitude and frequency content.
- Efficient sampling of such data requires adaptive strategies, like increasing rates during activity bursts.
Purpose of the Study:
- To develop a real-time, adaptive algorithm for selecting sampling rates in body sensor networks.
- To reduce the number of measured samples while preserving essential information from physiological data.
- To improve the efficiency of data acquisition in health-care and remote surveillance applications.
Main Methods:
- A neural network-based algorithm predicts subsequent samples and their uncertainties.
- Measurements are taken only when the prediction risk exceeds a selectable threshold.
- The algorithm adaptively adjusts the sampling rate in real-time.
Main Results:
- The method was applied to electromyogram, electrocardiogram, electroencephalogram, and body acceleration data.
- Sampling rates were reduced below the Nyquist limit, preserving accurate data representation and power spectral densities (PSD).
- Errors in signal estimation were around 10% at 60% of the Nyquist frequency, with good PSD representation.
Conclusions:
- The proposed method enables exceeding the Nyquist limit for non-stationary biomedical signals while retaining information.
- Applications in body sensor networks can reduce wireless communication and memory requirements, saving sensor power.
- This approach offers a more efficient way to sample and process physiological data from wearable sensors.
