Classification of drug tablets using hyperspectral imaging and wavelength selection with a GAWLS method modified for...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Accurate identification and classification of pharmaceutical tablets are critical for drug safety and efficacy.
• Traditional methods may lack the precision required for complex authentication and quality control.
• Hyperspectral imaging offers a non-destructive, high-throughput approach to analyzing sample composition.

Purpose of the Study:

• To develop and validate a hyperspectral imaging-based method for accurate drug tablet classification.
• To integrate the k-nearest neighbor (kNN) algorithm with spectral data for tablet identification.
• To propose and evaluate a genetic algorithm-based approach for selecting optimal wavelength regions for classification.

Main Methods:

• Application of hyperspectral imaging to capture infrared spectra across a 2D plane of drug tablets.
• Utilizing the k-nearest neighbor (kNN) algorithm for tablet classification based on spectral databases.
• Implementing a genetic algorithm for wavelength selection to identify crucial spectral regions for classification.

Main Results:

• The combined hyperspectral imaging and kNN method achieved correct overall tablet classification.
• Despite potential point-wise misclassifications, the method demonstrated robustness due to multi-point spectral analysis.
• The genetic algorithm successfully identified relevant wavelength regions, enhancing classification performance.

Conclusions:

• Hyperspectral imaging coupled with kNN provides a reliable method for drug tablet classification.
• Wavelength selection using genetic algorithms optimizes the classification process by focusing on informative spectral regions.
• The proposed approach offers a promising tool for pharmaceutical quality control and counterfeit detection.