Label-Free Identification of White Blood Cells Using Machine Learning

Area of Science:

• Biomedical Engineering
• Computational Biology
• Hematology

Background:

• Traditional white blood cell (WBC) differential counting relies on fluorescent markers and flow cytometry, requiring extensive sample preparation that can disturb cells.
• Current methods necessitate multiple steps and can impact cell integrity, limiting diagnostic potential.

Purpose of the Study:

• To develop and validate a novel, label-free method for classifying live, unstained WBCs using an imaging flow cytometer and machine learning.
• To demonstrate the feasibility of distinguishing WBC subtypes, including B and T lymphocytes, without fluorescent labeling.

Main Methods:

• Utilized an imaging flow cytometer to capture morphological data from live, unstained white blood cells.
• Applied machine learning algorithms for the classification and differentiation of WBCs and their subtypes.
• Validated the method using unstained samples from 85 donors.

Main Results:

• Achieved an average F1-score of 97% for overall WBC classification.
• Successfully distinguished between B and T lymphocytes with an average F1-score of 78%, a significant advancement for unlabeled samples.
• Demonstrated robust and highly accurate identification of WBCs, minimizing cellular disturbance.

Conclusions:

• The novel label-free approach offers a robust, accurate, and less disruptive method for WBC analysis.
• This technique enhances diagnostic capabilities, particularly for liquid biopsy applications, by leveraging cell morphology.
• The open-source workflow facilitates broader adoption and further research in label-free cell analysis.