Design of a complex bioimpedance spectrometer using DFT and undersampling for neural networks diagnostics

Area of Science:

• Biomedical Engineering
• Electrical Engineering
• Medical Diagnostics

Background:

• Electrical impedance spectroscopy (EIS) is a non-invasive, low-cost technique with significant medical applications.
• Its diagnostic potential, particularly in oncology, is an area of active research.
• Accurate and reliable bioimpedance measurement is crucial for clinical translation.

Purpose of the Study:

• To design and experimentally evaluate a multifrequencial complex bioimpedance analyzer.
• To assess the accuracy of the developed analyzer against commercial equipment.
• To explore the feasibility of using bioimpedance data for head and neck cancer detection with neural networks.

Main Methods:

• Development of a multifrequencial complex bioimpedance analyzer.
• Calculation of impedance amplitude and phase using Discrete Fourier Transform (DFT).
• Measurement of high-frequency signals using undersampling, covering a frequency range of 50 Hz to 500 kHz.

Main Results:

• The prototype accurately measured impedance values from 1 Ω to 50 kΩ.
• Accuracy analysis on passive components showed a mean error of 2.9% for magnitude and 0.69 degrees for phase.
• A neural network model utilizing bioimpedance, gender, age, and BMI achieved 77.5% accuracy in simulating two types of head and neck cancer diagnoses.

Conclusions:

• The developed bioimpedance analyzer demonstrates high accuracy and a wide measurement range, suitable for medical applications.
• The initial findings suggest that bioimpedance spectroscopy, combined with machine learning, holds potential for non-invasive cancer diagnostics.
• Further research is warranted to refine the neural network model and validate its clinical efficacy for head and neck cancer detection.