Identification of Gas Mixture Components with Multichannel Hierarchical Analysis of Time-Resolved Hyperspectral Data

Area of Science:

• Chemical sensing
• Biosensors
• Spectroscopy

Background:

• Chemical vapor sensors are crucial for medical diagnostics and environmental monitoring.
• Accurate identification of mixed gases remains a challenge for current sensor technologies, including electronic noses.
• Previous work introduced multichannel hierarchical analysis for time-resolved hyperspectral systems to improve spectral ambiguity.

Purpose of the Study:

• To demonstrate the identification of mixed gas components using time-resolved line hyperspectral measurements.
• To utilize genetically engineered M13 bacteriophages as gas-selective colorimetric sensors in an eight-sensor array.
• To develop an efficient machine learning classification method for complex gas mixtures.

Main Methods:

• Employed time-resolved line hyperspectral measurements with a bacteriophage-based colorimetric sensor array.
• Converted time-dependent spectral variations into a hyperspectral 3D data cube.
• Processed the data cube into a multichannel spectrogram using dimensionality reduction and a block average filter, followed by convolution filtering for hierarchical analysis.

Main Results:

• Achieved a classification accuracy of 93.9% for identifying pure and mixed gases.
• Successfully identified low concentrations (2 ppm) of acetone, ethanol, and xylene.
• Demonstrated effective capture of complex gas-induced spectral patterns and temporal dynamics.

Conclusions:

• The developed hyperspectral imaging and machine learning approach significantly improves mixed gas identification accuracy.
• Genetically engineered bacteriophages show promise as sensitive and selective colorimetric sensors for vapor detection.
• This technology holds potential for noninvasive disease diagnosis and advanced environmental monitoring.