26.3 A Closed-Loop Neuromodulation Chipset with 2-Level Classification Achieving 1.5Vpp CM Interference Tolerance, 35dB Stimulation Artifact Rejection in 0.5ms and 97.8 Sensitivity Seizure Detection

摘要

In closed-loop neuromodulators for epilepsy patients, nonidealities such as common-mode interference (CMI), stimulation artifacts (SA), electrode DC offset (DCO) and 1/f noise bring challenges for the sensing circuit to capture EEG signals (<1 mV) and for the backend classifier to detect patient-specific seizures accurately for precise stimulation. Chopper-stabilized capacitive-coupled instrument amplifiers (CS-CCIA) are widely used to suppress DCO and 1/f noise [1]–[3], and feedforward CM cancelling (CMC) is further added in [2] but still shows limited CMI tolerance (< 650mpp). The SA rejection in [4] requires 3s response time, which is too long for real-time SA rejection. Also, the accurate classifier in [5] operates all the time, resulting in huge digital power (674µW). This work presents an 8-channel closed-loop neuromodulation chipset with 2-level (coars + fine) classification. Applied in a deep-brain stimulation (DBS) system, in vivo measurement verifies that the 2-level classification scheme results in: (1) 35dB SA rejection in 0.5ms; (2) intermittent operation of the fine classifier, achieving ~1.16µW average power for classification. A feedback CMC (FB-CMC) is proposed for the CS-CCIA, achieving CMI tolerance up to 1. 5Vpp.