Biomedical circuits and systems

Biomedical circuit and system address interdisciplinary areas of Circuits and Systems and Life Sciences. Microelectronic is heavily used in life sciences, physical sciences and engineering. Figure 1 shows an example of System on a Chip (SoC) that designed for biomedical purpose (seizure prediction). The SoC is used in biomedical area but built with board knowledge of algorithm, analog and digital circuit. The SoC is fabricated in advanced silicon technology (Figure 1a) which consists of both dedicated analog and digital function blocks (Figure 1b). The function of the system is to predict seizure before it happens (Figure 1c), and the performance of the system is defined by certain metrics (Figure 1d). CENBRAIN is such a place that bridge between a wide variety of related areas and microelectronics.

Figure 1. Typical seizure prediction system: (a) A highly integrated seizure prediction system includes a bio-interface to collect data such as EEG, ECoG or via a MEA and an integrated circuit for analog and digital processing; (b) The hardware building blocks of the integrated circuit; (c) The functionality of the circuits is to classify different seizure stages based on the captured bio-signal; and (d) The performance of system is characterized by various metrics such as receiver operating characteristic (ROC) and area under curve (AUC) which are characterized by the true positive rate (TPR), false positive rate (FPR) of the system.

Analog Design for Bio-signal Acquisition.

Electroencephalography (EEG), electrocorticography (ECoG) and other technique that employ implantable microelectrode array (MEA) are electrophysiological monitoring methods that can record electrical signals of the brain (Figure 2). EEG consists of electrical signals from the scalp with noninvasive electrodes. ECoG reads signals from sensors placed either above or below the dura mater. Microelectrodes implanted at the cortical layers are used to record the local-field potential (LFP). These signals are direct cause of neuron activities, hence they can be used for various brain status monitoring and classification applications such as fatigue detection, seizure detection and seizure prediction.


Figure 2. Different signal acquisition methods and their characteristics: (a) brain anatomy; (b) advantage and limitations comparison; and (c) visualization of the advantage and limitations.

In CENBRAIN, we use various analog design techniques to acquire high-quality bio-signals with different characteristics. Moreover, we improve the safety, comfortability, lifetime of these circuits by eliminating or reduce the invasiveness, minimizing the power consumption, and enabling energy harvesting.

Algorithm and Digital Design for Bio-signal Processing

Machine learning algorithms such as support vector machine (SVM), deep neural network (DNN) and other classification methods can recognize certain pattern from high dimensional data. They can be used to detect or even predict some symptom based on the analysis of some bio-signals. For example, many publications have shown that DNN are effective to predict and detect seizure based on EEG signals. However, these algorithms are mostly executed by a desktop computer at current stage. For real daily usages, a dedicated low-power mobile processing apparatus or platform is required.

In CENBRAIN, we optimize SVM or DNN algorithms to reduce computation, memory requirement while preserving model accuracy. According to the algorithm model, new hardware and chips will be designed. The algorithm, architecture, circuit co-design methodology used in CENBRAIN allows the processing system to have state-of-the-art real-time and low power performance.