Automated detection of circulating tumor cells with naive Bayesian classifiers

Area of Science:

• Oncology
• Biomedical Engineering
• Machine Learning

Background:

• Personalized medicine relies on early disease detection and informed decisions.
• Circulating tumor cells (CTCs) are critical prognostic markers in cancer progression and treatment.
• Reliable, automated diagnostic tools are essential for quantifying CTCs.

Purpose of the Study:

• To develop an automated machine learning method for detecting and enumerating circulating tumor cells (CTCs).
• To utilize a naive Bayesian classifier (NBC) for CTC classification based on fluorescence signatures.
• To demonstrate the effectiveness of unsupervised learning for CTC detection using unlabeled data.

Main Methods:

• Cells were collected using a functionalized medical wire.
• Fluorescence microscopy and RGB color histograms were employed to capture cell color signatures.
• A naive Bayesian classifier (NBC) was implemented for automated cell classification.

Main Results:

• The NBC method successfully detected and enumerated CTCs based on their fluorescence signatures.
• The approach enabled unsupervised learning, eliminating the need for labeled training data.
• Performance was quantitatively compared and found competitive with state-of-the-art support vector machines.

Conclusions:

• The developed NBC method offers a reliable and automated approach for CTC detection and enumeration.
• This technique advances liquid biopsy applications in personalized medicine for cancer prognosis.
• The unsupervised learning capability makes the method adaptable to various unlabeled datasets.