Research on Intelligent Diagnosis Method of Swallowing Signal Based on Complex Electrical Impedance Myography

Swallowing is a complex neuromuscular reflex involving coordinated contraction and relaxation of throat muscles. These muscle activities cause changes in electrical impedance in the larynx, which can be measured to assess swallowing.

Swallowing disorders (dysphagia) are increasingly common in patients with stroke, neurological diseases, and in senior citizens, often leading to serious conditions like aspiration pneumonia and malnutrition. Accurate and early detection of swallowing function is crucial for diagnosing dysphagia, reducing complications, and improving patient health for a better quality of life.

Electrical Impedance Myography (EIM) is a promising technique that detects changes in muscle resistance, or impedance, as muscles move. However, traditional EIM systems capture only amplitude data and analyze signals using Fourier transforms, which can suffer from spectrum leakage due to incomplete signal cycles, limiting diagnostic accuracy.

The researchers developed a novel system called Complex Electrical Impedance Myography (C-EIM) based on an advanced method—Integer Period Digital Lock-In Amplifier (IPD-LIA) to monitor swallowing. IPD-LIA improves signal detection by capturing both amplitude and phase information with high speed and low noise.

The system has four key parts: a Field Programmable Gate Array (FPGA), high-speed analog-to-digital converters (ADCs), a custom analog front end (AFE), and four electrodes. Two electrodes deliver a tiny current through the neck muscles, while the other two measure the resulting voltage changes. The setup captures signals at high speed and with low noise, making it reliable for tracking the electrical properties of muscle movement in real-time.

For experimental validation, subjects performed various swallowing tasks—including dry swallowing, drinking water, and eating yogurt—while electrodes on both sides of the neck recorded muscle activity over 15 minutes. The data was analyzed by examining both amplitude and phase information from each swallow, assigning a weightage of 60% to amplitude and 40% to phase. The analysis revealed a strong correlation (r = 0.83) between amplitude and phase information as important markers of swallowing events.

An smart signal analysis based on the Genetic Algorithm-Generalized Regression Neural Network (GA-GRNN) was utilized to recognize different types of swallows. The GA  helps optimize the neural network to perform better. GRNN is especially useful when the data sample size is small and can classify both simple and complex patterns. Classification performance was evaluated through precision, recall, F1-score, and confusion matrices to ensure accuracy and reliability.

Using traditional EIM with amplitude-only data, the system achieved 89.3% accuracy and struggled with events like dry swallowing and water intake. In contrast, the enhanced C-EIM combined with GA-GRNN and both amplitude and phase data reached 95.2% accuracy. Several events, including yogurt and biscuit consumption and speech, exceeded 95% recognition rates, with some reaching 100%. F1 scores consistently surpassed 90%, confirming the model’s reliability. These results highlight the potential of C-EIM combined with smart algorithms as a robust, non-invasive tool for early dysphagia detection and improved clinical decision-making.

The proposed diagnosis method will significantly enhance the monitoring and identification of swallowing events. It will pave the way for more effective, accurate, and early intervention in dysphagia management.