Automated Detection and Cropping of Hyphae in Microscopic Images of Various Fungi

Date of Award

12-2021

Document Type

Thesis

Degree Name

Master of Science in Computer Science

First Advisor

Patricia Angela R. Abu, PhD

Abstract

Fusarium oxysporum f. sp. cubense is a soil-borne fungi that has become a major threat to the current banana industry. The presence of this fungi can destroy entire plantations if not detected and stopped early enough. The purpose of this study is to be able to create a Convolutional Neural Network (CNN) that can detect hyphae in microscopic images. By detecting hyphae, the presence of fungi in the soil can be confirmed. This study can act as a stepping stone for future studies involving the classifica- tion of fungi through hyphae and other features.

To be able to create a model that can detect hyphae, a dataset of var- ious microscopic images of fungi was sorted into hyphae images and non- hyphae images (labeled as others). Four subsequent datasets were be cre- ated from this, namely: (1) bright field, (2) dark field, (3) fluorescent, and (4) all microscopy techniques. Five folds cross validation using ResNet 34 and ResNet152 models was used for each of the datasets. The models were all able to achieve acceptable accuracies in predicting what images have hy- phae or not. After comparing the results, the models trained on fluorescent images were determined to be the ones that performed the best. To further improve the accuracies, an automated cropping method was used to crop out unnecessary noise and leave the hyphae of the image as the main sub- ject. Using the cropped images combined with the raw images for training a ResNet152 model resulted in higher accuracies for the fluorescent model.

Link to Full Text

Share

COinS