Identification and Staging of B-Cell Acute Lymphoblastic Leukemia Using Quantitative Phase Imaging and Machine...

Area of Science:

• Biomedical optics
• Computational pathology
• Machine learning in diagnostics

Background:

• Accurate and rapid identification of leukemia cells is crucial for timely diagnosis and treatment.
• Current diagnostic methods can be time-consuming and require specialized reagents.
• Quantitative phase imaging (QPI) offers label-free, high-content cellular analysis.

Purpose of the Study:

• To develop and evaluate machine learning (ML) classifiers for distinguishing healthy B cells from lymphoblasts.
• To classify stages of B-cell acute lymphoblastic leukemia (B-ALL) using QPI data.
• To assess the efficiency of ML models in terms of computational resources.

Main Methods:

• Utilized quantitative phase imaging to acquire label-free images of cells.
• Developed and trained machine learning classifiers, including convolutional neural networks (CNNs).
• Analyzed morphological parameters such as dry mass and volume for cell classification.

Main Results:

• Demonstrated that normal B cells have lower average dry mass and volume compared to cancerous cells.
• Observed an increase in these morphological parameters with disease progression in B-ALL.
• ML classifiers achieved effective cell type discrimination with minimal training requirements (space, time, memory).

Conclusions:

• QPI combined with ML provides a rapid, label-free method for leukemia cell classification.
• Morphological parameters derived from QPI are sensitive indicators of B-ALL and disease stage.
• This approach holds promise for objective, low-preparation hematopathology diagnostics, warranting further clinical studies.