Label-free tumor cells classification using deep learning and high-content imaging.

Area of Science:

• Biomedical Engineering
• Computational Biology
• Oncology

Background:

• Cellular morphology analysis is crucial for detecting circulating tumor cells (CTCs) in blood samples.
• Existing methods often lack robustness due to limited validation on heterogeneous cancer cell lines and non-blood normal cells.
• There is a need for advanced computational tools to accurately differentiate cancer cells from normal cells based on morphology.

Purpose of the Study:

• To develop and validate a deep learning model for distinguishing organoid-derived cancer cells from normal cells using diverse morphological features.
• To create a comprehensive dataset of microscopic images capturing morphological heterogeneity in cancer and normal cells.
• To establish a foundation for an automated platform for CTC detection.

Main Methods:

• Construction of a large-scale dataset comprising over 75,000 images of organoid-derived cancer and normal cells from three cholangiocarcinoma patients.
• Development of a proof-of-concept deep learning model for classifying cancer cells versus normal cells in unlabeled microscopy images.
• Validation of the model's performance using the area under the receiver operating characteristics curve (AUROC) metric.

Main Results:

• The deep learning model achieved an AUROC of 0.78 in distinguishing cancer cells from normal cells.
• The model demonstrated generalization capabilities, performing well on cell images from an unseen patient.
• The developed dataset provides a valuable resource for future research in CTC detection.

Conclusions:

• The study presents a novel deep learning approach for accurate cancer cell identification based on morphology.
• The model's performance indicates its potential for improving the sensitivity and specificity of CTC detection platforms.
• This work lays the groundwork for developing automated, robust systems for clinical applications in cancer diagnostics.