Deep Learning-Based Image Analysis for the Identification of 33 Bacterial Species: Convolutional Neural Network with Metadata Integration

Background: Accurate and timely identification of microorganisms is essential for clinical decision-making, biosafety assurance, and microbiological research. Traditional culture-based or biochemical approaches are time-consuming, highlighting the need for automated, image-based pipelines capable of species-level resolution. Methods: We developed a bacterial classification framework that integrates convolutional neural network (CNN) analysis of Gram-stained microscopy images with metadata-driven contextual reporting. A dataset of 2,034 images from 33 clinically and industrially relevant bacterial species was used. Images were preprocessed through normalization, resizing, and real-time augmentation before being fed into a sequential CNN optimized for hierarchical feature extraction. Additionally, a curated metadata repository covering eight laboratory-relevant attributes (e.g., Gram stain result, morphology, oxygen requirement, biosafety level, and pathogenicity) was incorporated to enhance interpretability and reporting. Results: On an independent test set, the model achieved an accuracy of 0.84, a weighted F1-score of 0.84, and a Matthews correlation coefficient of 0.84, indicating balanced classification across species. Distinct species such as Actinomyces israelii, Candida albicans, and Neisseria gonorrhoeae were identified with perfect precision and recall. In contrast, species with high intra-genus similarity, particularly within the Lactobacillus genus, exhibited comparatively lower performance. Conclusion: The integration of CNN-based image analysis with metadata enrichment enables the generation of rapid, standardized, and information-rich laboratory reports. This framework demonstrates potential for application in diagnostic, research, and educational settings, and represents a step toward fully integrated, multimodal microbiological identification pipelines suitable for future large-scale and real-world clinical deployment.