Neural network based AI model for lung health assessment

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Healthcare
• Pulmonology

Background:

• Pulmonary diseases represent a significant global health burden, being the third leading cause of mortality worldwide.
• Accurate and timely diagnosis of lung conditions is crucial for effective patient treatment and management.
• Current diagnostic methods can be enhanced by leveraging advanced technologies like artificial intelligence (AI).

Purpose of the Study:

• To develop and evaluate a novel artificial intelligence (AI) model for the analysis of lung sounds.
• To assess the diagnostic performance of the proposed AI approach using multiple public datasets.
• To demonstrate the generalizability and superiority of the AI model compared to existing methods for pulmonary disease diagnosis.

Main Methods:

• Utilized four datasets, combining two public sources, for comprehensive model assessment.
• Applied signal pre-processing techniques including normalization, re-sampling, and framing to lung sound recordings.
• Employed eight sub-band filters for frequency band segregation and extracted signal characteristics (entropy, L1 norm, kurtosis, etc.).
• Developed a neural network (NN) architecture with three fully connected layers and an output layer for classification.

Main Results:

• The proposed AI approach achieved 100% accuracy, specificity, and sensitivity across all four datasets.
• The model demonstrated strong generalizability, performing consistently well on diverse data.
• The NN architecture is characterized by its simplicity, ease of implementation, and short training duration.

Conclusions:

• The developed AI model offers a highly accurate and reliable method for diagnosing pulmonary diseases through lung sound analysis.
• The proposed neural network architecture significantly outperforms existing methods in terms of classification accuracy and efficiency.
• This AI-driven approach holds promise for improving diagnostic capabilities in respiratory medicine and reducing mortality rates associated with lung diseases.