Interphase Cell Cycle Staging using Deep Learning^.

Area of Science:

• Cell Biology
• Computational Biology
• Medical Imaging

Background:

• Cell cycle progression is vital for tissue homeostasis.
• Disruptions in cell cycle regulation are linked to diseases like cancer.
• Current automated tools for cell cycle staging, especially during interphase, are insufficient.

Purpose of the Study:

• To investigate deep learning approaches for interphase cell cycle staging in microscopy images.
• To develop a reliable automated tool for single-cell level classification.
• To enhance cancer prognosis and therapeutic strategy development.

Main Methods:

• Three deep learning approaches were evaluated: joint detection/classification, sequential detection then classification, and detection/segmentation then classification.
• Methods were applied to DAPI-stained nuclei in microscopy images.
• Approach 3, involving nuclei segmentation, achieved the best performance.

Main Results:

• The best performing method achieved an F1-Score of 0.908.
• Nuclei segmentation enabled intensity normalization, considering relative intensities within an image.
• Relative nuclear intensities are important for accurate cell cycle staging.

Conclusions:

• A novel deep learning method for interphase cell cycle staging at the single-cell level was developed.
• The method's success highlights the importance of considering relative nuclear intensities.
• This approach has potential implications for cancer prognosis and treatment strategies.