Detection of Breath Nitric Oxide at Ppb Level Based on Multiperiodic Spectral Reconstruction Neural Network

Area of Science:

• Analytical Chemistry
• Biomedical Engineering
• Optical Sensing

Background:

• Breath nitric oxide (NO) is a key biomarker for respiratory inflammation.
• Noninvasive diagnostic methods for respiratory conditions are crucial.
• Accurate detection of parts-per-billion (ppb)-level NO in exhaled breath is challenging.

Purpose of the Study:

• To develop a novel sensor for online detection of ppb-level NO in exhaled breath.
• To improve the accuracy and reliability of noninvasive respiratory inflammation diagnosis.
• To establish a robust optical method for breath analysis.

Main Methods:

• Developed a spectral reconstruction method to enhance NO absorption signals in the intensity domain.
• Utilized a multiperiodic spectral reconstruction neural network for NO concentration detection.
• Processed spectral data to remove noise and interference, improving target gas extraction.

Main Results:

• Achieved online detection of NO in the range of 1.63–846.68 ppb.
• Demonstrated high accuracy with a mean absolute error (MAE) of 0.31 ppb and a mean absolute percentage error (MAPE) of 0.96%.
• Successfully distinguished breath NO levels between healthy individuals and simulated patients in exhalation experiments.

Conclusions:

• The developed breath NO sensor offers a reliable and accurate noninvasive method for diagnosing respiratory inflammation.
• The spectral reconstruction and neural network approach effectively enhances NO detection sensitivity and specificity.
• This optical-based breath analysis technology holds significant potential for medical diagnostics.