A Comparative Study of Different Machine Learning Based Feature Extraction Techniques in Bacterial Image Classification

Abstract

Bacteria are single-celled, microscopic organisms that have the tendency to survive in a variety of conditions. Majority of these organisms found in soil, ocean, and many of them are also present in human body. There are few species of bacteria that are beneficial to humans, whereas most of them are adverse in nature and are responsible for contamination that causes variety of infectious diseases that includes Cholera, Strep throat, Tuberculosis, etc. As a result, precise bacterial identification and classification is critical in preventing the spread of such life-threatening pandemic. Traditional methodologies-based bacteria image classification and identification techniques are time consuming, less accurate and needs extremely high skilled microbiologists to deal with complex nature of above said problem. With the evolution and penetration of machine learning based computer assisted technologies in this field, various flaws and issues can be addressed easily. The model developed using machine learning tools and technologies are highly successful in this domain of image analysis and have shown extremely high rate of improvement in clinical microbiology investigation by identifying different bacteria species. To improve upon results, feature extraction from digital images is essential and extremely important for better and accurate classification of bacteria. Feature extraction aids in removing unnecessary data from a data set and adds to the increase in speed of learning and generalization in entire machine learning process. This study presents a comparative study of research undertaken using different machine learning based techniques in feature extraction relating digital bacterial images that leads to effective and efficient classification and identification of different species. The study also identified and recommends the suitable classifier which is best at giving results using different feature extraction methods.