Defense Date

6-29-2020

Graduation Date

Summer 8-8-2020

Availability

One-year Embargo

Submission Type

thesis

Degree Name

MS

Department

Biomedical Engineering

School

Rangos School of Health Sciences

Committee Chair

Melikhan Tanyeri

Committee Member

Kimberly Forsten Williams

Keywords

Droplet Microfluidics, Machine Learning, Support Vector Machines, E. Coli, Bacteria Quantification and Detection, Droplet Classification, Intensity Based Algorithm, Bacteria Images

Abstract

Sepsis is a major medical problem and massive resources have been invested in developing and evaluating alternative treatments. Statistics indicate that sepsis causes between one third and one half of all hospital deaths in the United States. Sepsis has a high impact on health care in the US, with direct sepsis costs in 2009 exceeding $15.4 billion. A research study found that a 1-hour delay in appropriate antimicrobial care resulted in a 7% - 10% rise in mortality. Several professional societies seek to reduce sepsis mortality by targeting the timely use of diagnostic tests and antimicrobial therapy. The diagnostic instruments available to clinicians to identify the suspected pathogen do not make a timely intervention possible. Up to 5 days of incubation are needed for blood cultures, the majority of bacteria being detected after 12–48 h. Therefore, fast and simple techniques are required for rapid bacterial cell detection and quantification. By using droplet microfluidics and a machine learning algorithm, the objective of this study was to propose a technology that analyzes images of bacterial cells by image processing and Support Vector Machines algorithm to classify droplets containing the bacteria. The accuracy of the proposed technology was 97.2 % for a trained SVM model and with the complete identification and classification of droplets.

Language

English

Download

Share

COinS