Skip to main content
2024 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

BMES High School Student Chapter Member

HS Poster

Poster 8 - AI/ML Based Virtual Reality Digital Twin for Biofilm Formation and their Response to Intervention

Saturday, October 26, 2024
10:00 AM - 12:00 PM EST
Location: Exhibit Hall E, F & G

Introduction: Biofilm formation significantly impacts medicine, environment, food safety, industry, and biotechnology, influencing issues such as antibiotic resistance, corrosion, waste treatment, and foodborne diseases. Traditional 2D visualization methods offer limited insights, highlighting a need for comprehensive VR-compatible 3D reconstructions. This study addresses this gap by creating a digital twin of biofilm formation and intervention, transforming 2D stack images into 3D models using AI and machine learning for VR applications.

Materials and

Methods: Initially, the optical density (OD) growth readings were collected from available literature and experiments. This data provides information on the density of microbial/biofilm growth over time. A database of OD values was created, encompassing various experimental conditions and microbial strains. This database serves as the foundation for modeling biofilm growth and responses to chemical interventions. Logistic models were applied to fit growth curves, offering a mathematical representation of microbial growth dynamics. The logistic model generates factors that describe how microbial density changes under different conditions. These factors are integrated into a microbial growth simulation, which dynamically updates the digital twin of the biofilm. Initial microbial community configurations were derived from image data to ensure accurate representation of starting conditions. The communities were represented by spheres, with their sizes reflecting the density of the biofilms. The simulation was finally visualized in 3D using matplotlib. To achieve a realistic experience in a VR environment, the Segment Anything Model (SAM) was used to segment microbial regions from the 2D stack images, replacing the Otsu thresholding method from our previous approach. This new method enhanced the accuracy and detail of biofilm structure segmentation, which is crucial for creating fine-grained 3D models. The segmented biofilm images were then reconstructed to generate the actual 3D assets for visualization in the VR environment.

Results, Conclusions, and Discussions: Figure 2 shows the modeling result and the growth simulation side by side, and the biofilm in the VR environment. The red dots show the actual OD readings, and the blue line shows the predicted values from the logistic model. Our findings confirm that a digital twin of biofilm offers a promising tool for in-depth biofilm analysis in medical research and environmental sciences.

Our machine learning-based digital twin effectively reconstructs 3D biofilm structures from 2D microscopy images, allowing for immersive visualization in VR environments. The proposed digital twin can simulate biofilm growth under various chemical interventions, integrating image analysis, deep learning, computer vision, medicine, and microbiology to provide an enriched, unbiased view of biofilm structures and their interactions with anti-biofilm compounds. Future endeavors will focus on refining extending the image database for various bacteria.

Acknowledgements (Optional):

References (Optional):