Channel Parameter Estimation in Neural Communication Based on Bullfrog Sciatic Nerve

Advances in molecular communication and the Internet of Nanothings (IoNT) have opened new possibilities for in-body nanodevice networks in medicine. As a promising approach for transmitting information from IoNT to external devices, neural communication leverages the nervous system as a data transmission interface. However, reliable transmission depends on accurate channel parameter estimation and research in this area remains limited. To address this, we take pH as the primary channel parameter and experimentally propose channel parameter estimation schemes using the bullfrog sciatic nerve as the neural communication channel. Here, compound action potentials (CAPs), which are the electrical responses of the nerve channel, are employed to characterize the channel. To establish the relationship between CAPs and pH values, we develop an experimental platform to collect CAPs at different pH values. Then, we design regression models including the random forest (RF) and long short-term memory (LSTM) methods, and further propose an enhanced LSTM model to address their limitations. The enhanced model uses convolutional layers to extract local spatial features from raw CAPs waveforms, followed by LSTM layers for temporal modeling, and concludes with MLP layers for continuous pH prediction. Experimental results reveal that the corresponding pH value can be accurately detected when new CAPs are fed into the trained models, with the enhanced LSTM model demonstrating superior accuracy. This study paves the way for conducting experimental research to ensure reliable data transmission from in-vivo IoNT to external networks.