Automatic processing and real-time analysis system for deep brain electrical signals based on sensing-enabled DBS

Deep brain stimulation (DBS), as an established neuromodulation technique, has become an important therapeutic approach for neuropsychiatric disorders such as Parkinson’s disease and obsessive-compulsive disorder. In recent years, sensing-enabled DBS devices have achieved a significant breakthrough from unidirectional “brain writing” to bidirectional “brain reading-writing” by recording local field potentials (LFPs) while delivering stimulation, advancing the development of bidirectional closed-loop brain-machine integration. However, LFP signal research based on sensing-enabled DBS faces two core scientific challenges: first, the signal-noise separation challenge—how to effectively separate target physiological signals from noise in recorded signals within the fully implanted environment; second, the collaborative optimization of real-time performance and safety—how to achieve real-time precise analysis and output modulation while ensuring secure brain-machine data interaction. To address these challenges, this study proposes systematic methods for deep brain electrical signal processing and analysis based on sensing-enabled DBS. For the complexity of noise artifacts, we designed an automatic identification algorithm based on a dual-branch deep neural network, achieving 93.9% identification accuracy on the validation set and 86.0% accuracy on cross-patient and cross-annotator test sets, establishing an automated analysis pipeline from preprocessing to feature extraction. For real-time and safety issues, we proposed a real-time signal transmission and online processing method based on secure protocols, with an average signal transmission latency increment of only 1.32±0.37 ms. Building upon this, we further integrated the NERCN NeuroSignal Research (NNSR) platform, achieving comprehensive end-to-end technical support for deep brain signals acquired from sensing-enabled DBS, spanning from offline analysis to online real-time transmission and processing. Validation demonstrates that this research can identify deep brain neural activity features in complex environments and support real-time analysis, providing reliable technical support for exploring neuromodulation mechanisms and therapeutic innovation.