Deep Learning-Based Object Detection And Bacteria Morphological Feature Extraction For Antibiotic Mode Of Action Study
Issued Date
2023-01-01
Resource Type
Scopus ID
2-s2.0-85179554249
Journal Title
BMEiCON 2023 - 15th Biomedical Engineering International Conference
Rights Holder(s)
SCOPUS
Bibliographic Citation
BMEiCON 2023 - 15th Biomedical Engineering International Conference (2023)
Suggested Citation
Chotayapa K., Leethamchayo T., Chinnawong P., Samernate T., Nonejuie P., Achakulvisut T. Deep Learning-Based Object Detection And Bacteria Morphological Feature Extraction For Antibiotic Mode Of Action Study. BMEiCON 2023 - 15th Biomedical Engineering International Conference (2023). doi:10.1109/BMEiCON60347.2023.10322010 Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/91561
Title
Deep Learning-Based Object Detection And Bacteria Morphological Feature Extraction For Antibiotic Mode Of Action Study
Author's Affiliation
Other Contributor(s)
Abstract
The increasing prevalence of antimicrobial resistance (AMR), as microorganisms develop resistance to antimicrobial drugs, has emerged as a critical concern in infection treatment, resulting in a rising death toll. Assessing the effect of drugs can provide insights by studying the morphological change of bacteria after drug treatment. However, utilizing conventional techniques such as CellProfiler for long-term and large-scale sample experiments is impractical due to the manual processes involved. To address this challenge, we proposed a deep learning-based object detection model for predicting the type of antibiotic treatment and automatically extracting bacteria morphology. Our model combines YOLOX and two Cascade R-CNNs using weight box fusion to enhance performance. It achieves an mIOU of 0.753 and mAP of 0.699 higher mAP compared to CellProfiler (mAP = 0.218). In addition, we use a computer vision approach to extract bacteria morphological features including cell membrane, DNA, and color intensity to classify the treated antibiotic which achieves comparable performance to CellProfiler (F1-Score = 0.75, 0.79 respectively). We believe our work can be used as an automatic tool to enhance the efficiency of antibiotic prediction and extracting cell profiles for AMR applications. Our code and web application are available at https://github.com/biodatlab/bacteria-detection.