Week 9: Research Proposal

 Hello! 

                                                                      

                                                                                 this week's samples :)

    This week, our research proposal was due, and on the assignment, it was recommended to post what we wrote here. However, as a quick update, we had to throw out the samples we did before spring break since they grew way too much. As there are three of us on this project, we have decided to delegate tasks by bacteria class, and each of us is assigned to our own class. I chose gram-positive bacilli. I started with the gram-positive bacilli samples today and will take pictures tomorrow, Wednesday, and next Monday. I made a total of 9 plates since, in my category, there are 3 different types of bacteria, and we will need to dispose of each sample that we take a photo of since we need to remove the lid. We want to avoid the samples getting contaminated. Anyways, here is what I wrote for my research proposal:

Research Proposal: Machine Learning and Bacteria Classification

Our research proposal is to develop a machine learning algorithm that accurately determines a bacteria species by its characteristics measured from high-quality still images. This topic is significant as identifying and classifying bacteria is a critical task in microbiology since it enables early disease detection, diagnosis, and treatment. When bacteria classification is done through traditional methods, it can take up to 24-48 hours to receive results (Wang et al., 2020). Additionally, current identification methods, such as culturing and biochemical tests, are time-consuming and may not accurately identify specific bacterial species in multispecies samples. Thus, there is a need to develop more efficient and accurate methods for bacterial classification and recognition.

Learning more about the intersection of deep learning algorithms and bacteria is essential as it could potentially reduce infection and improve efficiency regarding early detection models. This improvement would be beneficial as, although only 1% of known bacteria species are harmful to human health (Gallardo-García et al., 2022), the treatment of waterborne diseases costs 2 billion USD in the United States alone, with a recorded 90 million cases each year (Wang et al., 2020). One of the bacteria that impact human health is Escherichia coli (E. coli), which is responsible for 265,000 cases in the United States on average (NC Department of Health and Human Services, 2019).

Expanding the current knowledge in machine learning-based bacterial classification and recognition is crucial since it has the potential to revolutionize the diagnosis and treatment of bacterial infections. By developing this model, there is potential to expand this research into bacteria movement behavior, as Ma et al. (2020) researched, and even further uncover other antibiotic structures and underlying biological processes for bacteria and disease evolution (Jiang et al., 2022). By improving the accuracy and efficiency of bacterial identification, clinicians can provide timely and appropriate treatment, reducing the morbidity and mortality associated with bacterial infections. Additionally, developing more accurate and efficient methods for bacterial classification can aid in surveilling bacterial outbreaks, enabling a prompt public health response to prevent the spread of infectious diseases.

We will be researching our topic by creating nine separate algorithms for three different classes of bacteria. We have determined this will be the simplest way to provide a proof of concept for our specific needs and materials. The bacteria we will be using are as follows:

Gram-negative Bacilli

Enterobacter aerogenes (EA) 

Escherichia coli (EC)

Pseudomonas aeruginosa (PA)

Gram-positive Bacilli

Bacillus subtilis (BS)

Corynebacteria xerosis (CX)

Bacillus cereus (BC)

Gram-positive Cocci

Staphylococcus epidermidis (SE)

Staphylococcus aureus (SA)

Enterococcus faecalis (EF)

We will first sample each bacteria from each class and take photos at 24 hours, 48 hours, and one week. Each time will have its own bacteria sample so as not to contaminate the species during the time we take the photo and during incubation. In each photo, we will aim to image a single colony so that the algorithm can measure and characterize each colony. After this process, we will determine its accuracy when given an image from that same bacteria and use two controls to compare against: a positive control in which a bacteria outside of its class is presented, and a negative control, an empty plate with no particulates. After we have successful accuracies of at least 90% from each of our nine algorithms, we will attempt to combine each algorithm into a single class (creating a total of 3 algorithms) and, in the future, one comprehensive algorithm. 

Our research goal is to achieve a deep-learning algorithm that can determine a bacteria species with an accuracy of at least 90%. The goal will be achieved by feeding multiple high-quality images of each bacteria species in its respective class, giving us higher accuracy results and a foundation for building a comprehensive singular program. Results we may discover are lower or higher accuracy predictions and if feeding additional photos of each bacteria type will increase the program's confidence in determining bacteria without classification parameters initially set. 

References

Gallardo García, R., Jarquín Rodríguez, S., Beltrán Martínez, B., Hernández Gracidas, C., & Martínez Torres, R. (2022). Efficient deep learning architectures for fast identification of bacterial strains in resource-constrained devices. Multimedia Tools and Applications, 81(28), 39915–39944. https://doi.org/10.1007/s11042-022-13022-8 

Jiang, Y., Luo, J., Huang, D., Liu, Y., & Li, D.-dan. (2022). Machine learning advances in microbiology: A review of methods and applications. Frontiers in Microbiology, 13. https://doi.org/10.3389/fmicb.2022.925454 

Ma, Y. (2020). Classification of bacterial motility using machine learning. University of Tennessee. 

NC Department of Health and Human Services. (2019, December 16). Escherichia coli (E. coli) Infection. NCDHHS. Retrieved March 17, 2023, from https://epi.dph.ncdhhs.gov/cd/diseases/ecoli.html#:~:text=Each%20year%20in%20the%20United,illnesses%20and%20about%20100%20deaths. 

Wang, H., Ceylan Koydemir, H., Qiu, Y., Bai, B., Zhang, Y., Jin, Y., Tok, S., Yilmaz, E. C., Gumustekin, E., Rivenson, Y., & Ozcan, A. (2020). Early detection and classification of live bacteria using time-lapse coherent imaging and Deep Learning. Light: Science & Applications, 9(1). https://doi.org/10.1038/s41377-020-00358-9