OoCount: A Machine-Learning Based Approach to Mouse Ovarian Follicle Counting and Classification

Area of Science:

• Reproductive biology and developmental science.
• Bioimaging and microscopy.
• Computational biology and machine learning.

Background:

• Accurate ovarian follicle counts are crucial for assessing ovarian health and function.
• Traditional methods for counting oocytes in 3D ovarian images are time-consuming and labor-intensive.
• Deep learning offers potential for rapid, automated analysis of microscopy images.

Approach:

• Developed OoCount, an open-source, high-throughput method for automatic oocyte segmentation and classification.
• Utilized a convolutional neural network (CNN) within DL4MicEverywhere for oocyte labeling.
• Employed Accelerated Pixel and Object Classification for sorting oocytes into growth stages.
• Established a fast tissue-clearing and spinning disk confocal imaging protocol for whole mouse ovaries.

Key Points:

• OoCount enables automatic segmentation and classification of oocytes from 3D fluorescent microscopy images of whole mouse ovaries.
• The protocol includes tissue clearing, 3D imaging, and machine learning-based analysis for follicle counting and staging.
• OoCount is customizable for images generated in different laboratory settings.
• Achieved accurate counts of oocytes across various growth stages in perinatal and adult mouse ovaries.

Conclusions:

• OoCount provides an efficient and accurate method for quantifying ovarian follicles.
• This tool enhances the study of ovarian function, reproductive health, and fertility.
• The open-source nature of OoCount promotes accessibility and customization for researchers worldwide.