Processing code-multiplexed Coulter signals via deep convolutional neural networks

Area of Science:

• Biotechnology
• Microfluidics
• Signal Processing

Background:

• Coulter sensors traditionally count and size particles.
• Microfluidic integration of multiple Coulter sensors allows spatial particle tracking.
• Code-multiplexing simplifies hardware but requires advanced signal processing for multi-dimensional data extraction.

Purpose of the Study:

• To integrate deep learning-based signal analysis with microfluidic code-multiplexed Coulter sensor networks.
• To train convolutional neural networks for analyzing Coulter waveforms and resolving interferences.
• To predict particle size, speed, and location using advanced signal processing.

Main Methods:

• Coupling deep learning (convolutional neural networks) with microfluidic code-multiplexed Coulter sensor networks.
• Training neural networks to recognize sensor waveform patterns and resolve interferences.
• Utilizing advanced signal processing to extract multi-dimensional information from output waveforms.

Main Results:

• Achieved >90% pattern recognition accuracy for distinguishing non-correlated waveform patterns.
• Demonstrated potential for real-time microfluidic assays due to processing speed.
• Showcased the algorithm's ability to predict particle size, speed, and location.

Conclusions:

• Deep learning-based signal analysis enhances microfluidic Coulter sensor networks for particle tracking.
• The developed algorithm offers high accuracy and potential for real-time applications.
• The trained algorithm is adaptable for processing data from other microfluidic devices with similar sensor networks.