Myelin detection in fluorescence microscopy images using machine learning.

Area of Science:

• Neuroscience
• Biomedical Imaging
• Computational Biology

Background:

• Myelin sheath, produced by glial cells, is crucial for nervous system function.
• Degeneration of myelin causes neurodegenerative disorders such as multiple sclerosis (MS).
• Current MS therapies lack remyelination-promoting treatments, and drug discovery is hindered by manual myelin quantification.

Purpose of the Study:

• To develop an automated, machine learning-based approach for expedited myelin detection in fluorescence microscopy images.
• To facilitate high-content, image-based drug screening for remyelination therapies in MS.

Main Methods:

• Developed a machine learning model for myelin detection in 3D fluorescence microscopy images.
• Utilized a spectro-spatial feature extraction method for voxel analysis.
• Trained and evaluated 23 supervised machine learning techniques, including a customized convolutional neural network (CNN), on over 47,000 annotated images.

Main Results:

• Achieved high accuracy in myelin detection, with Boosted Trees reaching 98.84±0.09% and a customized CNN reaching 98.46±0.11%.
• The approach successfully segments myelin in 3D from multi-channel z-stack fluorescence images, regardless of orientation.
• Demonstrated robustness in a common experimental setup.

Conclusions:

• The proposed expedited myelin detection method is a feasible and robust tool for remyelination drug screening.
• This automated approach can significantly accelerate the drug discovery process for MS and other myelin-related disorders.
• Enables efficient, high-throughput screening for compounds that promote myelin repair.