DeepLearnMOR: a deep-learning framework for fluorescence image-based classification of organelle morphology.

Area of Science:

• Plant cell biology
• Computational biology
• Bioimage analysis

Background:

• Organelle biogenesis, morphogenesis, and dynamics are crucial for cellular function.
• Current methods for assessing organelle morphology are manual, slow, and require expertise.
• Deep learning offers potential for high-throughput, automated image analysis.

Purpose of the Study:

• To develop and validate a deep learning framework for analyzing organelle morphology in plants.
• To automate the classification and identification of morphological abnormalities in energy organelles.
• To enhance the speed and efficiency of plant cell imaging screens.

Main Methods:

• Utilized transfer learning and a convolutional neural network (CNN).
• Analyzed over 47,000 confocal microscopy images of Arabidopsis thaliana.
• Focused on three key energy organelles: chloroplasts, mitochondria, and peroxisomes.

Main Results:

• Developed the DeepLearnMOR (Deep Learning of the Morphology of Organelles) framework.
• Achieved over 97% accuracy in classifying image categories and identifying morphological abnormalities.
• Feature visualization confirmed CNN's decision-making process, ensuring reliability and interpretability.

Conclusions:

• The DeepLearnMOR framework provides a rapid and accurate method for organelle morphology analysis.
• This approach significantly improves the efficiency of image-based screens for plant organelle research.
• Establishes a foundation for future large-scale studies with diverse datasets.