Automated bone cell classification for confocal laser scanning microscopy volumes.

Area of Science:

• Biomedical Engineering
• Cell Biology
• Microscopy Imaging

Background:

• Manual cell classification in microscopy is subjective and time-consuming.
• Identifying bone cells is challenging due to irregular shapes and bone mineral presence.
• Fluorescence-based lineage tracing increases subjectivity in cell classification.

Purpose of the Study:

• To develop and validate an automated algorithm for objective cell identification and classification in 3D bone cell cultures.
• To classify osteoblast-lineage cells based on cell body color in fluorescence microscopy images.
• To compare the automated algorithm's performance against manual labeling.

Main Methods:

• Developed a MATLAB-based automated algorithm for cell identification and classification.
• Utilized flattened volumetric image stacks from 3D bone cell cultures.
• Validated the algorithm by comparing its output to manual labeling on eleven images.

Main Results:

• The automated algorithm achieved precision, recall, and F1 scores higher than 0.75 for all cell classifications, with most exceeding 0.80.
• No significant differences were observed between manual and automated cell counts or classifications.
• Automated analysis averaged 17 seconds per image, compared to over 10 minutes for manual labeling.

Conclusions:

• The developed algorithm provides a fast, repeatable, and accurate method for classifying bone cells using fluorescence in confocal microscopy.
• This automated process can be applied to other pre-clinical models and pathological tissues for cell identification.
• The algorithm objectively addresses the limitations of manual cell classification in complex biological samples.