SYSTEMS AND METHODS FOR THE DETECTION AND CLASSIFICATION OF LIVE MICROORGANISMS USING THIN FILM TRANSISTOR (TFT) IMAGE SENSOR AND DEEP LEARNING

Smart overview of the Invention

The patent application describes a bacterial colony-forming-unit (CFU) detection system that utilizes a thin-film-transistor (TFT)-based image sensor array. This system is designed to significantly reduce the time required for detecting and classifying live microorganisms compared to traditional Environmental Protection Agency (EPA)-approved methods. The integration of advanced imaging technology and deep learning algorithms enables rapid and accurate analysis of bacterial colonies.

Imaging Modality

A lensfree imaging modality is employed, leveraging the TFT image sensor to capture a sample field-of-view of approximately 10 cm². Time-lapse images are automatically collected at 5-minute intervals, allowing for continuous monitoring of bacterial growth. This approach facilitates the observation of dynamic changes in the colonies over time, enhancing the detection process.

Deep Learning Integration

The system incorporates two deep neural networks to perform critical tasks: detecting and counting growing bacterial colonies and identifying their species. These networks are trained to recognize specific patterns and features associated with different microorganisms, ensuring high accuracy in classification. The use of deep learning significantly enhances the system's ability to process and analyze complex image data.

Performance Evaluation

In blind tests involving 265 colonies of E. coli and other coliform bacteria, such as Citrobacter and Klebsiella pneumoniae, the system demonstrated impressive performance metrics. It achieved an average CFU detection rate of 97.3% within 9 hours of incubation and an average recovery rate of 91.6% at approximately 12 hours. These results highlight the system's efficiency and reliability in real-world applications.

Scalability and Applications

The TFT-based sensor technology is adaptable to various microbiological detection methods, offering versatility across different research and clinical settings. Additionally, the imaging field-of-view can be cost-effectively expanded to cover areas greater than 100 cm², making it suitable for large-scale applications. This scalability ensures that the system can meet diverse needs while maintaining its cost-effectiveness.