Classification of chemically modified red blood cells in microflow using machine learning video analysis

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cell Biology

Background:

• Traditional cytometers often rely on immunophenotyping, which can be complex and require fluorescent labels.
• Analyzing red blood cell (RBC) morphology and dynamics is crucial for diagnosing various pathological conditions.
• There is a need for rapid, label-free methods to categorize RBCs.

Purpose of the Study:

• To develop and validate an AI-based video classifier for distinguishing native and chemically modified red blood cells.
• To assess the capability of AI video analysis in capturing RBC morphology and motion dynamics.
• To demonstrate a novel microfluidic approach for rapid, label-free cell categorization.

Main Methods:

• Utilized TensorFlow for video analysis of red blood cells.
• Employed a microfluidic cytometer to capture cell shape and motion trajectories.
• Chemically modified RBCs in three distinct ways to simulate pathological conditions.
• Trained an AI video classifier to differentiate between native and modified RBCs.

Main Results:

• Achieved classification accuracies exceeding 90% for all three types of modified cells compared to native cells.
• Demonstrated that AI video analysis effectively captures RBC morphology and motion dynamics.
• The microfluidic cytometer successfully categorized cells based on shape and flow without fluorescence labels.

Conclusions:

• AI-based video classification provides a highly accurate and efficient method for categorizing red blood cells.
• This label-free, microfluidic approach offers a significant advancement over traditional immunophenotyping cytometers.
• The technology holds potential for rapid diagnosis of RBC-related pathological conditions.