Cell Group Recognition Method Based on Adaptive Mutation PSO-SVM.

Area of Science:

• Biomedical Detection Technology
• Computational Biology
• Statistical Analysis

Background:

• High-throughput flow cytometry (FCM) generates large, complex multidimensional data, increasing demand for robust statistical analysis methods.
• Support Vector Machines (SVM) are effective for classification, but their clustering accuracy heavily depends on optimal selection of penalty factor (c) and kernel parameter (g).
• Traditional Particle Swarm Optimization (PSO) can struggle with local optima and poor convergence for large datasets.

Purpose of the Study:

• To develop an optimized SVM algorithm for improved FCM data clustering.
• To address the parameter optimization challenges in SVM for accurate cell identification.
• To enhance the reliability and efficiency of multidimensional FCM data analysis.

Main Methods:

• Proposed an adaptive mutation particle swarm optimization (PSO) algorithm integrated with SVM (PSO-SVM) for parameter optimization.
• Selected an appropriate kernel function for the FCM sample data.
• Applied the adaptive mutation PSO algorithm to optimize SVM parameters for cell clustering.
• Validated the method using extensive FCM experimental data.

Main Results:

• The proposed PSO-SVM method significantly improved clustering correctness by 19.38% compared to traditional SVM.
• Achieved a high clustering correctness rate of 99.79% for FCM data.
• Demonstrated superior performance over cross-validation and standard PSO algorithms in FCM data clustering.
• Showcased lower time complexity and reduced need for manual intervention.

Conclusions:

• The adaptive mutation PSO-SVM algorithm offers a highly accurate and efficient solution for analyzing complex FCM data.
• This method effectively overcomes the limitations of traditional SVM and PSO, particularly for large-scale biological datasets.
• The developed approach advances cell group identification in biomedical detection technology.