Classification of fungal genera from microscopic images using artificial intelligence.

Area of Science:

• Medical Mycology
• Computational Biology
• Machine Learning

Background:

• Microscopic examination is a cornerstone of clinical microbiology for diagnosing fungal infections.
• Accurate and rapid identification of pathogenic fungi is crucial for effective patient management.
• Traditional methods can be time-consuming and require specialized expertise.

Purpose of the Study:

• To develop and evaluate deep convolutional neural networks (CNNs) for classifying pathogenic fungi from microscopic images.
• To compare the performance of various CNN architectures in fungal species identification.
• To assess the potential of a deep learning approach to enhance fungal diagnostics.

Main Methods:

• Trained multiple CNN architectures (DenseNet, Inception ResNet, InceptionV3, Xception, ResNet50, VGG16, VGG19) on a dataset of 1079 microscopic fungal images from 89 genera.
• Data was split into training, validation, and test sets at a 7:1:2 ratio.
• Performance was evaluated using top 1 and top 3 prediction accuracy; further analysis included excluding rare genera and applying data augmentation.

Main Results:

• The DenseNet CNN model achieved the highest performance, with 65.35% top 1 and 75.19% top 3 accuracy for classifying 89 fungal genera.
• Performance improved to over 80% accuracy after excluding rare genera and implementing data augmentation.
• 100% prediction accuracy was achieved for certain specific fungal genera.

Conclusions:

• Deep learning, particularly using the DenseNet CNN model, offers a promising automated approach for identifying filamentous fungi from microscopic images.
• This method has the potential to significantly enhance diagnostic accuracy and reduce the turnaround time for fungal identification in clinical settings.
• The findings suggest a valuable tool for augmenting traditional mycological diagnostic workflows.