26.2 A Neuromorphic Multiplier-Less Bit-Serial Weight-Memory-Optimized 1024-Tree Brain-State Classifier and Neuromodulation SoC with an 8-Channel Noise-Shaping SAR ADC Array
Gerard O’Leary, Jianxiong Xu, Liam Long, José Sales Filho, Camilo Tejeiro, Maged ElAnsary, Chenxi Tang, Homeira Moradi, Prajay Shah, Taufik A. Valiante, Roman Genov
Abstract
Personalized medical brain implants have the potential to revolutionize the treatment of neurological disorders and augment cognition. Critically, these devices require accurate, energy-efficient brain-state classifiers to determine the precise moment when the treatment neuromodulation efficacy is maximized, such as before the onset of a seizure in epilepsy [1]. The SoC presented in this work addresses this requirement by combining a bank of 8 neural signal ADCs with BrainForest, an accurate, low-power classification core comprised of a 1024-tree exponentially decaying memory decision forest (EDM-DF). Full closed-loop neuromodulation is supported through the responsive actuation of an on-chip electrical neurostimulator.