Medical condition diagnosis is heavily based on sensor measurements and their processing. These measurements correspond to waveforms that propagate over the complex body environment and are transformed linearly or nonlinearly according to the characteristic environment properties. However, the medical community does not fully exploit the potentials of advanced processing matched to nonlinear structures or modern sensor technologies such as waveform agility that leads to significant estimation performance improvements. This research exploits advanced implementation-aware sensing and processing techniques to improve medical diagnosis by: (a) efficient processing using compressed sensing and nonlinear time-varying spectral methods; (b) estimation of environment descriptors and disease state parameters combined with waveform-agile sensing; and (c) mapping estimation and waveform-agile sensing algorithms onto field-programmable gate arrays. This framework brings revolutionary advances in diagnosing, treating, and tracking disease states that are otherwise difficult to obtain as advanced processing techniques are either not available or too costly. The investigators also design on-line software toolboxes with sensing experiments for use in outreach programs to recruit and retain freshmen and underrepresented student populations.
The research integrates advanced signal processing, stochastic Bayesian estimation, and waveform-agile sensing with implementation-aware algorithms to improve estimation of disease states. The investigators study time-frequency techniques matched to nonlinear structures to process biomedical data. Compressive sensing methodologies are developed that reduce the number of required measurements and allow for alternative computational algorithms. Mathematical descriptors are designed to model disease states using time-varying transfer functions. Sequential Bayesian techniques and waveform adaptation are used to estimate information. Algorithm reconstruction procedures are developed that trade high performance for reduced implementation cost. Finally, the investigators design algorithms cognizant of complexity and memory requirements.
