From Physiological Signals to Medical Information: Systems for Extracting Value from Low-power Sensors
Speaker: Naveen Verma, Princeton University
Department: Electrical Engineering
Location: Engineering Quadrangle J201
Date/Time: Thursday, April 11, 2013, 7:30 p.m. - 9:00 p.m.
Extremely promising modalities have emerged both for chronically sensing and stimulating human physiology. These could form the basis for devices capable of providing unprecedented medical decision support, therapeutic value, and/or prosthetic functions. The problem is that advanced devices to really exploit these modalities within clinical applications require the ability to convert physical sensor signals into specific, clinically?relevant information . Unfortunately, the underlying physiology of interest and the corresponding signals we can sense are extremely complex to model on an analytical level, especially to the high?order required for engineering practical clinical devices. Data?driven methods raise an alternate approach to modeling physiological signals. These use physical waveforms as the starting point, and they enable the creation of accurate models. Recent advances from the domain of machine learning have led to powerful frameworks for data?driven modeling and analysis. These can provide a means to exploit the large amount of data available in the medical domain (i.e., both retrospectively in databases and prospectively via chronic sensors) for high?order analysis of physiological signals acquired from sensors . The critical challenge is that traditional DSP platforms cannot adequately support the computational frameworks involved under the low?power system constraints faced by wearable and implantable devices. This talk describes hardware and algorithmic approaches that target these challenges. Several hardware prototypes are described that can address a wide range of clinical applications while reducing the computational energy, memory requirements, and signal?bandwidth impact by orders of magnitude [3,4].
 A. Csavoy, et al., Creating support circuits for the nervous system: Considerations for brain?machine interfacing, Symp. on VLSI Circuits (VLSI), June 2009.
 N. Verma, K. H. Lee, A. Shoeb, Data?driven Approaches for Computation in Intelligent Biomedical Devices: A Case Study of EEG Monitoring for Chronic Seizure Detection, Journal of Low Power Electronics and Applications, April 2011.
 N. Verma, et al., "A Micro?Power EEG Acquisition SoC With Integrated Feature Extraction Processor for a Chronic Seizure Detection System," IEEE Journal of Solid?State Circuits (JSSC), April 2010.
 K. H. Lee and N. Verma, A 1.2?0.55V General?purpose Biomedical Processor with Configurable Machine?learning Accelerators for High?order, Patient?adaptive Monitoring, European Solid State Circuits Conf. (ESSCIRC), Sept 2012.