Off-campus SJSU users: To download campus access theses, please use the following link to log into our proxy server with your SJSU library user name and PIN.

Publication Date

Fall 2020

Degree Type

Thesis - Campus Access Only

Degree Name

Master of Science (MS)

Department

Biomedical Engineering

Advisor

Melinda Simon

Keywords

artificial neural network, COMSOL, deterministic lateral displacement, electronic nose, urinary tract infection, volatile organic compounds

Subject Areas

Biomedical engineering

Abstract

An electronic nose technology was developed that aimed at an in vitro classification of urinary samples infected with different strains of Escherichia coli to predict the causing agents of urinary tract infection. The predictive model was optimized by developing a culturing protocol to ensure consistent assay as well as utilizing a headspace GC/MS method to characterize ethanol content in the samples. By employing a robust artificial neural network, it was shown to be able to extract the unique patterns of VOCs of the samples and successfully differentiate between those classes. For datasets collected from multiple measurements, the prediction accuracy was improved by including training data from the appropriate first few measurement cycles, which were deemed quality data from a stability test with the GC/MS. Lastly, a passive cell separation technique called deterministic lateral displacement (DLD) was enhanced by applying gap size variation while keeping the row shift fraction unchanged. This novel approach, which increases separation resolution while maintaining the microsystem throughput, was found to be disjunctive from the conventional method for not being predictable by the standard critical size equation. Future research could explore the capability of the DLD system as a bacterial concentrator and the possibility of automation, which highlights its great potentials for incorporating into the e-nose system.

Download

Share

COinS