Multi-modal brain monitoring systems for healthcare applications

Abstract

Multimodal measurement of the brain activity has an obvious advantage compared with unimodal measurement. Multimodal analysis has two advantages. First, the more information about brain activity can be achieved. Second, with the combination of the classifier, the effect of artifacts that are reflected in only one modality can be reduced. Therefore, the multimodal measurement turns out to have the improved classification accuracy. In the first SoC, a multimodal mental management system in the shape of the wearable headband and earplugs is proposed to monitor electroencephalography (EEG), hemoencephalography (HEG) and heart rate variability (HRV) for accurate mental health monitoring. It enables simultaneous transcranial electrical stimulation (tES) together with real-time monitoring. The total weight of the proposed system is less than 200 g. The multi-loop low-noise amplifier (MLLNA) achieves over 130 dB CMRR for EEG sensing and the capacitive correlated-double sampling transimpedance amplifier (CCTIA) has low-noise characteristics for HEG and HRV sensing. Measured three-physiology domains such as neural, vascular and autonomic domain signals are combined with canonical correlation analysis (CCA) and temporal kernel canonical correlation analysis (tkCCA) algorithm to find the neural-vascular-autonomic coupling. It supports highly accurate classification with the 19% maximum improvement with multimodal monitoring. For the multi-channel stimulation functionality, after-effects maximization monitoring and sympathetic nerve disorder monitoring, the stimulator is designed as reconfigurable. The 3.37 × 2.25 $mm^2$ chip has 2-channel EEG sensor front-end, 2-channel NIRS sensor front-end, NIRS current driver to drive dual-wavelength VCSEL and 6-b DAC current source for tES mode. It dissipates 24 mW with 2 mA stimulation current and 5 mA NIRS driver current. In the second SoC, a multimodal head-patch system measuring both EEG and NIRS on the frontal lobe simultaneously is proposed for accurate anesthesia depth monitoring. With the help of MM-DSL, the LNA shows the state-of-the-art NEF of 3.59 at the 300mV EDO input. Moreover, impedance boosting loop enhances the input impedance up to 1G$omega$. Logarithmic TIA can reject ambient light of 10p-10nA to maximize the dynamic range up to 60dB. According to the comparator-based settling monitor output, NIRS driver duty-cycle can be adjusted from 0.625m-50ms adaptively. Extracted features are processed with deep neural network (DNN) to show the anesthesia depth index. As a result, the compact anesthesia depth monitoring head-patch enables more accurate anesthesia depth monitoring even under special drugs which the BIS cannot detect the anesthesia for safe surgery in operating room. With the proposed system, clinical results for propofol-induced general anesthesia and ketamine-induced general anesthesia are shown, respectively. First, the estimated anesthesia depth values of the proposed system are compared with reference values, in this case BIS output, during the same surgical operation. Proposed depth index trends such as sudden drop after the propofol sedation and steady increase after the reduction of inhalational anesthetic are almost the same as the index of reference. During intubation, which causes intense EMG, and electrocautery step, sudden BIS index rising (10-15) is observed but the proposed system generates stable results. Second, ketamine, which BIS gives false result, is used to test the operation of the proposed system. Its output clearly shows the clinically important transition from the awake to deep state but BIS cannot detect the transition. These ICs are integrated into compact sensor system. Thanks to multimodal measurement of brain activity, more accurate and compact system can be implemented. The proposed systems are fully implemented and verified by in-vivo experiments.