Deep learning-based classification of microalgae using light and scanning electron microscopy images.

Area of Science:

• Marine Biology
• Biotechnology
• Computational Science

Background:

• Microalgae have vital industrial and ecological roles.
• Algal blooms pose significant environmental threats, necessitating effective monitoring.
• Current identification methods are time-consuming, labor-intensive, and costly.

Purpose of the Study:

• To evaluate deep learning models for microalgae identification using both LM and SEM.
• To compare the performance of various Convolutional Neural Network (CNN) architectures.
• To investigate the efficacy of incorporating SEM images alongside traditional LM images.

Main Methods:

• Utilized light microscopy (LM) and scanning electron microscopy (SEM) images of five microalgae species (two Cyanobacteria, three Chlorophyta).
• Applied state-of-the-art Convolutional Neural Network (CNN) models: VGG16, MobileNet V2, Xception, NasnetMobile, and EfficientNetV2.
• For the first time, integrated SEM images into deep learning-based microalgae identification.

Main Results:

• Achieved exceptional classification accuracy of 99% for both LM and SEM images with VGG16 and EfficientNetV2 models.
• Demonstrated high classification accuracies (>93%) for most models with SEM images.
• NasnetMobile showed the lowest accuracy (87%) with SEM images, while other models performed well.

Conclusions:

• Deep learning, particularly CNNs, offers a rapid and precise method for microalgae identification.
• Incorporating SEM images alongside LM images enhances identification accuracy.
• Cost-effective optical microscopy coupled with deep learning achieved superior results compared to electron microscopy for algal identification.