Counting bacterial colonies is a fundamental task in microbiology, which is currently performed manually in most laboratories. This is a time-consuming and error-prone process, which requires a trained professional.
As one of our internal R&D projects, partially funded by The National Centre for Research and Development, we work on a library for automatic identification and classification of bacterial colonies based on RGB images of Petri dishes. Our deep learning methods allow us to detect and count different types of microorganisms with high accuracy and can be easily integrated with lab automation software or used as a standalone application.
A flexible library for microbiological analyses will be created as a result of this R&D project. It is going to be easily customizable for new types of microorganisms, agar plates or camera parameters. It is intended to work in different scenarios:
– a standalone application for analysing already collected images;
– an application connected to a simple image acquisition set consisting of a camera mounted on a tripod (a standard set for a laboratory available on the market, such as the ones produced by Carl Zeiss);
– a library integrated with existing advanced systems for automating microbiological laboratory environment (i.e. MicroTechniX).
The areas of application include but not limited to healthcare, pharmaceutical, food, cosmetics, and veterinary industries.
To assure the high accuracy and generalization capabilities of developed algorithms the state-of-the-art deep learning methods are used, which are known to obtain the best results for the majority of computer vision tasks. More precisely, the solution is based on convolutional neural networks in a combination with suitable image processing pipeline to maximize the precision of detecting, classifying, and counting different species of microorganisms.
The crucial ingredient for deep learning-based solutions is data. That is why in collaboration with microbiology experts from the University of Wroclaw we are collecting thousands of carefully annotated images of Petri dishes with different bacterial species. By varying photos acquisition settings (e.g. lighting conditions, camera parameters etc.) we aim to make our algorithms robust to the variations in the image data collection process, so our solution can be applied to a variety of laboratory settings.
