Biology and Computation for Improving Access to Health Screening

Abstract

Many patients who need medical care are underserved by today’s systems, leaving them unable to get adequate warning when health conditions require medical attention. Computing technology can aid in reducing this healthcare gap by commoditizing sensors capable of providing diagnostic and prognostic information in an affordable and usable form factor. Ubiquitous sensing technologies, such as those built into modern smartphones, can be considered more accessible or affordable because the hardware can be re-used for many purposes, effectively amortizing the cost of each function over the full utility of the device. However, the generalized purpose designed into these sensors introduces unintended noise into signals which otherwise contain useful health information. Creatively designed data gathering techniques that leverage chemical and medical understanding of the underlying biology and physiology can detect signals that can be useful for health screening. In this thesis, I demonstrate these techniques in four projects over two areas: (1) blood-oxygen saturation and (2) hemoglobin concentration for health sensing in vivo using an unmodified smartphone camera, and (3) passively gathering samples of viral particles and (4) capacitance-based detection of specific nucleic acids for environment sensing in vitro. In order for technology to deliver on the promise to aid in healthcare, a physiological understanding of the molecular components can be combined with computational techniques, such as machine learning and signal processing, to reveal the underlying signals of interest, improving access to health screening for the growing proportion of the population with smartphones.