A novel hybrid approach for microbial detection using deep ensemble learning.

Area of Science:

• Microbiology and Biotechnology
• Computer Science and Artificial Intelligence

Background:

• Accurate microbial detection and classification are crucial in healthcare, clinical microbiology, biotechnology, and environmental health.
• Deep learning models have significantly improved image-based microbial analysis, enhancing accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate DeepBACT, a hybrid ensemble model for accurate microbial image detection and classification.
• To leverage transfer learning and feature selection for efficient microbial analysis.

Main Methods:

• A hybrid ensemble model, DeepBACT, was created by combining VGG19, ResNet50, and InceptionV3.
• Principal Component Analysis (PCA) was used for feature selection, followed by Support Vector Machine (SVM) classification.
• The model was trained and tested on the novel Bacterial Images (BACTI) dataset.

Main Results:

• DeepBACT achieved an F1-score of 97% and an accuracy of 98.61% on the BACTI dataset.
• The model demonstrated superior performance compared to state-of-the-art methods in both accuracy and efficiency.
• Evaluated using randomized train-test split and 5-cross validation.

Conclusions:

• DeepBACT effectively classifies unseen microbial images with high accuracy and efficiency.
• The proposed hybrid ensemble approach is suitable for real-time deployment in microbial detection and classification tasks.