Spatio-Temporal Features in Microbial Culture Recognition

• Grant number: 24/07102-8
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: August 01, 2024
• End date: June 30, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Antonio Rafael Sabino Parmezan
• Grantee: João Pedro Ribeiro da Silva
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 19/17721-9 - The role of Chemistry in holobiont adaptation, AP.TEM

Abstract

Many challenges faced in data mining and knowledge discovery are related to data classification. This task has been explored in various applications, such as microorganism identification based on image analysis of cultures grown in Petri dishes. Studies on this topic have demonstrated considerable predictive performance, with error rates below 15% for image collections containing more than five bacteria species. This project questions whether this is the case or if it is possible to further reduce the microbe classification error through video prediction techniques. Given that the culture images to be used in this research were captured over several days at equidistant intervals, we propose treating them as video frames to improve microorganism recognition. Several architectures have been developed for this purpose, some based on multi-label learning theory, while others rely on ensembles, meta-learning, and deep learning. Investigating these approaches, which involve spatial-temporal feature extraction, is the primary goal of this project. The built classification models will be evaluated mainly in terms of accuracy, precision, recall, and F1-score.