Deterministic Lateral Displacement (DLD) Analysis Tool Utilizing Machine Learning towards High-Throughput Separation.

Area of Science:

• Biomedical Engineering
• Microfluidics
• Computational Biology

Background:

• Deterministic lateral displacement (DLD) is a microfluidic technique for size-based particle separation.
• DLD shows promise for isolating circulating tumor cells (CTCs) in blood for cancer diagnostics.
• Current DLD data analysis is manual, time-consuming, and prone to errors.

Purpose of the Study:

• To develop an automated particle detection and data analysis method for DLD microfluidic devices.
• To improve the accuracy and reduce the processing time of DLD separation experiments.
• To enable high-throughput analysis for cancer diagnostics and therapeutics.

Main Methods:

• Utilized Python and machine vision techniques for particle detection.
• Implemented and compared three machine learning models for DLD separation mode determination.
• Developed a reliable particle detection algorithm for microfluidic analysis.

Main Results:

• Achieved an overall particle detection accuracy of 97.86%.
• Reduced video analysis time significantly.
• Demonstrated an average computation time of 25.274 seconds for analysis.

Conclusions:

• The developed automated method significantly reduces human error and processing time in DLD analysis.
• This approach supports the advancement of high-throughput DLD devices for cancer diagnostics.
• Automated data analysis is crucial for the clinical translation of microfluidic technologies.