Evaluation of flow-volume spirometric test using neural network based prediction and principal component analysis.

Area of Science:

• Pulmonary Medicine
• Biomedical Engineering
• Data Science

Background:

• Spirometric pulmonary function tests are crucial for diagnosing respiratory diseases.
• Enhancing the diagnostic accuracy of spirometry, especially with incomplete data, remains a challenge.

Purpose of the Study:

• To improve the diagnostic relevance of spirometric pulmonary function tests.
• To utilize neural networks and Principal Component Analysis (PCA) for enhanced spirometry analysis.

Main Methods:

• Generated 175 flow-volume curves using spirometers under a standard protocol.
• Employed a neural network (backpropagation) to predict the key parameter, Forced Expiratory Volume in 1 second (FEV(1)).
• Applied PCA to analyze parameter interdependencies in measured and predicted datasets.

Main Results:

• The neural network successfully predicted FEV(1) in both normal and abnormal spirometry cases.
• PCA results indicated that FEV(1) is a significant discriminator between normal and abnormal lung function datasets.
• The percentage variance in principal components confirmed FEV(1)'s discriminative power.

Conclusions:

• The combined approach of neural network prediction and PCA offers a valuable tool for spirometric pulmonary function testing.
• This method shows potential utility in analyzing spirometry data, particularly when dealing with incomplete datasets.
• The study highlights the effectiveness of AI and data analysis techniques in respiratory diagnostics.