Combining neural network and genetic algorithm for prediction of lung sounds

Area of Science:

• Pulmonary Medicine
• Biomedical Engineering
• Artificial Intelligence

Background:

• Accurate lung sound classification is crucial for diagnosing pulmonary diseases.
• Traditional methods may lack efficiency and accuracy in complex sound analysis.
• Automated systems can aid clinicians in objective lung sound interpretation.

Purpose of the Study:

• To develop and evaluate a hybrid neural network-genetic algorithm approach for automated lung sound classification.
• To improve the prediction accuracy of adventitious lung sounds (wheeze and crackle).
• To optimize neural network structure and training parameters for reduced computational load.

Main Methods:

• Lung sounds were recorded from healthy and diseased subjects.
• Fourier Power Spectrum Density (PSD) was calculated for breath cycles.
• Genetic algorithms selected optimal spectral analysis data (129 values).
• A Multilayer Perceptron (MLP) neural network trained with backpropagation was used for classification.

Main Results:

• The hybrid approach successfully identified adventitious lung sounds.
• Genetic algorithms optimized the neural network architecture and parameters.
• Reduced processing load and time were achieved through network optimization.

Conclusions:

• The proposed neural network-genetic algorithm model offers an efficient and accurate method for lung sound classification.
• This approach has the potential to enhance the diagnostic capabilities in pulmonary medicine.
• Optimization through genetic algorithms leads to more efficient AI models for medical signal processing.