Date of Award

January 2011

Document Type

Open Access Thesis

Degree Name

Medical Doctor (MD)

Department

Medicine

First Advisor

Sheldon M. Campbell

Subject Area(s)

Microbiology, Computer science

Abstract

Digital image analysis for the interpretation of images in clinical microbiology has many potential advantages over current practices. Compared to traditional image interpretation by a medical technologist, digital image analysis offers standardization between laboratories, round-the-clock interpretation, and quantitative results. In the first study of its kind known to the authors, a digital image analysis program was prototyped to interpret a slide containing Gram stained microorganisms. The sample microorganisms were obtained from culture plates during routine processing and subjected to Gram's stain. An initial study learned from 11 Gram-stained slides and classified their microorganisms into the group: Gram-positive, Gram-negative, rods, coccus, and yeast. The sensitivity of identification ranged from 66% to 99% and the specificity ranged from 78% to 99%. The algorithm was next applied to a larger set of 78 slides. The accuracy rate for slide classification was 60 out of 78 or 77%. After using this larger dataset to train the algorithm, the accuracy rate for individual objects was 94% averaged over 5 trials. This suggests the parameters used by the algorithm can differentiate between groups, and the lack of accuracy in classifying the larger database occurred due to limitations in the original training data. Overall, the project demonstrates a unique application of digital image analysis to clinical microbiology.

Comments

This is an Open Access Thesis.

Download
Open Access

This Article is Open Access

Share

COinS