Digital Signal Processing in Bio-implantable Systems: Design Challenges and Emerging Solutions Seetharam Narasimhan 1 , Jongsun Park 2 , Swarup Bhunia 1 1 Electrical Engineering and Computer Science Department, Case Western Reserve University, 10900 Euclid Avenue, Cleveland, OH USA 2 School of Electrical Engineering, Korea University, Seoul, Korea E-mail: sxn124@case.edu, jongsun@korea.ac.kr, skb21@case.edu Abstract Implantable systems that monitor biological signals require increasingly complex digital signal processing (DSP) electronics for real-time in-situ analysis and compression of the recorded signals. While it is well-known that such signal processing hardware needs to be implemented under tight area and power constraints for small footprint and increased battery-life, new design requirements emerge with their increasing complexity. Use of nanoscale technology shows tremendous benefits in implementing these advanced circuits due to dramatic improvement in integration density and power dissipation per operation. However, it also brings in new challenges such as reliability and high leakage power. Besides, programmability of the device and security of the recorded information are desirable features, which need to be considered during the design of such systems. Programmability is important to adapt to individual subjects as well as to the temporal fluctuations in subject condition. On the other hand, information security is rapidly becoming an important design parameter since the recorded signal often needs to be transmitted outside the body through wireless channels. In this paper, we analyze the emerging issues associated with the design of the DSP unit in an implantable system. We note that conventional design solutions may not be attractive for such systems. However, novel algorithm-architecture-circuit co-design solutions, which leverage on the nature of the signal processing algorithms can be effective to realize ultra low-power, robust, programmable and secure hardware for on-chip real- time signal processing in implantable systems. Keywords Bio-implantable systems, digital signal processing, neural interface, ultralow power design 1. Introduction With great advances in electronics and electrode technology, it has become possible to implement implantable systems, which interface with the biological organisms to monitor various biological signals and even manipulate the actions using electrical/chemical stimulation. One of the success stories in the field of biomedical devices is the cardiac pacemaker [1], which has been implanted in countless human beings. With numerous biomedical devices being used for interfacing with different body parts to save or enhance lives of millions, pervasive implantable devices are rapidly becoming a reality. Fig. 1(a) shows some example applications of bio-implantable devices. These devices are increasingly being used to recognize and treat symptoms of various diseases like epilepsy, heart disease, Parkinson’s disease, blindness, urinary incontinence etc. Researchers are also using the implantable devices as interfaces to the central nervous system to achieve better understanding of the mechanisms of neural communication and control. By studying simple organisms with tractable nervous systems, one can gain insight into the correlation between patterns of neural activity at the level of individual neurons and the resultant behavior of the organism [8]. Such behaviorally meaningful patterns can range from single spikes in a single neuron to timed bursts of neural spikes from a population of neurons, depending on the granularity of the behavior being studied. Numerous efforts have been made to use arrays of electrodes and associated electronics for understanding the signals in a complex nervous system [2]. Implantable neural interfaces have been explored in diverse contexts including neural stimulation, as in cardiac pacing and Functional Electrical Stimulation (FES). FES of nerves or muscles is used to assist patients in grasping, standing, or urination, while deep brain stimulation (DBS) has been shown to be an effective treatment for Parkinson’s disease. Cochlear implants are commercially available for treating deafness in children, while visual prostheses have had preliminary success in creating sensations of vision. Extensive research has been done on developing Brain Computer Interfaces (BCI) [3] in which a tetraplegic person can control movement of a computer cursor or a robotic arm. Current implementations of these systems, however, do not perform in-situ signal processing using digital circuits, although some of them use simple control algorithms based on external sensor data. The need for a closed-loop neural system, which records from multiple neurons, analyzes the neural activity and stimulates some neurons based on the analysis, has been emphasized before [4]. However, most of the current neural interface systems employ sophisticated data analysis performed on an external computer. Real-time closed-loop neural control can greatly benefit from in-situ signal processing using low-power miniaturized hardware. Such in-situ processing is more important for chronic implantations as well as to facilitate ambulatory movements of a patient. Although intense research has been carried out on designing the analog front-end circuitry [2, 5] as well as algorithms for off-line signal analysis, the design of algorithms and digital circuits for online signal processing inside the implantable system is comparatively new. In the context of neural signal processing, Harrison proposed a simple thresholding scheme for on-chip spike detection 978-1-4244-7808-8/10/$26.00 ©2010 IEEE 223 2nd Asia Symposium on Quality Electronic Design