Analyzing microscopic characteristics of biofilms via deep learning

Author: Rushil Gadamshetty, Computer Science and Data Driven Material Discovery Center for Bioengineering Innovation

Contributor: Md Hafizur Rahman, Department of Electrical Engineering

Mentor: Dr. Bharat Jasthi, Department of Materials and Metallurgical Engineering and BuG ReMeDEE Consortium

Mentor: Dr. Venkata Gadhamshetty, Department of Civil and Environmental Engineering, Data Driven Material Discovery Center for Bioengineering Innovation, and BuG ReMeDEE Consortium

Mentor: Dr. Shankarachary Ragi, Department of Electrical Engineering

Biofilms are a multicellular colony of microorganisms that can grow on virtually any surface. Although biofilms seem to barely interfere with human activities, they can be extremely beneficial and at times quite dangerous. In this research paper, we are analyzing the biofilm phenotype to better understand their behavior and prevent biofilm formation in certain areas. Our arching goal is to develop artificial intelligence (AI) approaches for developing a class of next generation conformal corrosion-resistant coatings for controlling biofilm growth. To extract quantitative data from microscopy images, we are implementing image processing and deep learning algorithms to automate the process of measuring microscopic characteristics of biofilms, which are otherwise laborious, time-consuming, and costly tasks. We implement a deep learning approach called Mask R-CNN to detect and segment the bacterial cells in the biofilm and subsequently measure the size characteristics of the cells. We present a case study based on biofilms of sulfate-reducing bacteria grown on metal surfaces intentionally modified with conformal coatings of two-dimensional (2D) materials. The practical implications of this study are to discover new material solutions for solving pressing biofilm problems including corrosion and fouling issues.