Cardiovascular disease and Diabetes Mellitus are chronic diseases that affect the quality of life of tens of millions of Americans and result in hundreds of billions of dollars in economic impact within the United States. Remote monitoring medical devices can provide real time updates of physiological parameters that have the potential to improve disease management, thereby enhancing life quality and reducing the financial burden. However, developing these medical devices to be cost effective and robust to patient-specific factors has proven difficult to this point. Computational modeling and simulation is a valuable tool which can aid in the process of designing medical devices and overcoming the barriers experienced to this point. Specifically in this work, computational modeling and simulation frameworks are developed to (1) enable design and evaluation of an insertable glucose biosensor and (2) to estimate the impact of patient and device specific factors on extracting photoplethysmographic waveforms, ultimately for use in remote blood pressure monitor. A computational framework for a multi-modal optical and fully-insertable glucose biosensor was developed via Monte Carlo modeling and Finite Element Method. The optical output of such a biosensor when implanted 2 mm in the volar wrist was determined by first validating the representation of a phosphorescence lifetime decay assay and a Forster Resonance Energy Transfer against known assays. It was found that using near infrared components yields sufficient signal to be detectable by a simple photodiode across skin tones. This framework was then used to determine biosensor geometries that would yield stronger luminescent output, and it was found that a stacked cylinder design 0.43 cm in length with 0.036 cm repeating units would yield maximum luminescent output and have limited chemical crosstalk. A combined Monte Carlo and gaussian combination framework was developed to estimate the impact of patient specific factors such as skin tone and age and device specific factors such as device wavelength on extracting Photoplethysmograph (PPG) waveforms for blood pressure (BP) prediction. In this work, Monte Carlo was used to estimate the signal strength of photoplethysmography under certain conditions, and gaussian combination was used to generate synthetic waveforms impacted by combinations of parameters. This framework was used to analyze the impact these factors have on signal strength, feature extraction, and the predictive precision of in-house neural network, bagged trees, and support vector machine algorithms. It was shown that patient specific factors such as age have a large effect on feature extraction, whereas patient specific factors such as skin tone do not. Additionally, differences in signal processing methods generated a large impact in extracting PPG for BP prediction. In this work, we create computational models that aid in the development of medical devices for remote monitoring of diabetes and cardiovascular disease. These models are validated, used to determine the feasibility of in-development medical devices, and are lastly used to demonstrate the impact of patient specific and device factors on device performance.
