Bacterial Behaviour Analysis Through Image Segmentation Using Deep Learning Approaches

TY - CONF

T1 - Bacterial Behaviour Analysis Through Image Segmentation Using Deep Learning Approaches

AU - Rahman, Afroza

AU - Rahman, Miraz

AU - Ahad, Md Atiqur Rahman

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

PY - 2024

Y1 - 2024

N2 - Antimicrobial Resistance (AMR) refers to the ability of microorganisms to resist the effects of certain medicines. Medicines that were previously known effective against diseases caused by different types of microorganisms are now incompetent towards the same treatment because of AMR, which also increases the risk of severe illness. By understanding AMR and the potential factors that lead to it, we can see how microorganism behaviour analysis has become a great tool. The limitation of human visual capabilities requires automated image-based solutions to analyse bacterial behaviour effectively. In this paper, we exploit growth stage-based multiple images of bacteria, i.e. E. coli (Escherichia coli) to Analyse bacterial behaviours to get valuable insight. We have used the Deep Learning algorithms to get segmented images for each of the growth stages. Our objective is to use U-net and StarDist to get bacterial behavioural features and compare their performances in terms of Ground Truth and predicted segmented masks. For both the Ground Truth and predicted segmented mask, we have determined total bacterial cell count, average bacteria volume, central distance from the image center, total area, average aspect, average solidity, average extent, average orientation, average Local Binary Patterns (LBP) and features of Gray-Level Co-occurrence Matrix (GLCM) such as contrast, dissimilarity, homogeneity, energy, and Angular Second Moment for each of the images. Also, we have analysed area change and movement from one frame to another frame, which represents bacterial growth over specific periods. Analysing these features will allow the researcher to identify the best-performing model for each of the calculating features of bacteria. Comparing these features between the actual mask and predicted segmented mask can help to identify valuable insights regarding bacterial behaviour which can be useful to identify factors that contribute towards AMR.

AB - Antimicrobial Resistance (AMR) refers to the ability of microorganisms to resist the effects of certain medicines. Medicines that were previously known effective against diseases caused by different types of microorganisms are now incompetent towards the same treatment because of AMR, which also increases the risk of severe illness. By understanding AMR and the potential factors that lead to it, we can see how microorganism behaviour analysis has become a great tool. The limitation of human visual capabilities requires automated image-based solutions to analyse bacterial behaviour effectively. In this paper, we exploit growth stage-based multiple images of bacteria, i.e. E. coli (Escherichia coli) to Analyse bacterial behaviours to get valuable insight. We have used the Deep Learning algorithms to get segmented images for each of the growth stages. Our objective is to use U-net and StarDist to get bacterial behavioural features and compare their performances in terms of Ground Truth and predicted segmented masks. For both the Ground Truth and predicted segmented mask, we have determined total bacterial cell count, average bacteria volume, central distance from the image center, total area, average aspect, average solidity, average extent, average orientation, average Local Binary Patterns (LBP) and features of Gray-Level Co-occurrence Matrix (GLCM) such as contrast, dissimilarity, homogeneity, energy, and Angular Second Moment for each of the images. Also, we have analysed area change and movement from one frame to another frame, which represents bacterial growth over specific periods. Analysing these features will allow the researcher to identify the best-performing model for each of the calculating features of bacteria. Comparing these features between the actual mask and predicted segmented mask can help to identify valuable insights regarding bacterial behaviour which can be useful to identify factors that contribute towards AMR.

KW - AMR

KW - Antimicrobial Resistance

KW - Bacterial Behaviour Analysis

KW - Deep Learning

KW - Image Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85202289941&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-67285-9_13

DO - 10.1007/978-3-031-67285-9_13

M3 - Paper

AN - SCOPUS:85202289941

SP - 172

EP - 185

T2 - 1st International Conference on Artificial Intelligence in Healthcare, AIiH 2024

Y2 - 4 September 2024 through 6 September 2024

ER -