Automated particle classification based on digital acquisition and analysis of flow cytometric pulse waveforms

Area of Science:

• Biomedical Engineering
• Computational Biology
• Analytical Chemistry

Background:

• Traditional flow cytometry analog processing limits waveform feature extraction to integral, height, and width.
• Advanced feature extraction is crucial for improved cell discrimination and analysis.

Purpose of the Study:

• To demonstrate the utility of Fourier properties of pulse waveforms for cell discrimination in flow cytometry.
• To implement and evaluate neural networks for automatic cell classification using digital flow cytometry data.
• To compare neural network performance against K-means clustering for cell classification.

Main Methods:

• Direct digitization of flow cytometry waveforms to extract features like skewness, kurtosis, and Fourier properties.
• Development and application of neural networks for automatic cell classification, including analysis without explicit feature extraction.
• Comparison of neural network classification performance with the K-means clustering algorithm.

Main Results:

• Fourier properties of pulse waveforms enable discrimination of cell types not separable by time-domain features alone.
• Neural networks provide efficient, automated cell classification without user interaction.
• Neural networks demonstrate comparable or superior performance to K-means clustering, with potential for real-time applications.

Conclusions:

• Digital flow cytometry with advanced waveform analysis, particularly Fourier properties, significantly enhances cell classification capabilities.
• Neural networks offer a powerful, automated, and potentially real-time solution for complex flow cytometry data analysis.
• The findings support the development of real-time digital data acquisition systems for advanced flow cytometry analysis.