Pharmaceutical Raw Material Identification Using Miniature Near-Infrared (MicroNIR) Spectroscopy and Supervised...

Area of Science:

• Analytical Chemistry
• Pharmaceutical Science
• Spectroscopy

Background:

• Pharmaceutical raw material identification (RMID) is critical for quality and safety.
• Near-infrared (NIR) spectroscopy offers rapid, non-destructive analysis for RMID.
• Challenges include model transferability and large-scale classification accuracy.

Purpose of the Study:

• To demonstrate the use of portable miniature NIR (MicroNIR) spectrometers for pharmaceutical RMID.
• To address model transferability and large-scale classification challenges using Support Vector Machine (SVM) modeling.
• To compare SVM performance against conventional chemometric methods.

Main Methods:

• Utilized six MicroNIR spectrometers and a set of 19 pharmaceutical compounds to test model transferability.
• Employed a dataset of 253 diverse pharmaceutical compounds for large-scale classification testing.
• Applied Support Vector Machine (SVM) modeling with a linear kernel, including a hierarchical scheme.
• Compared SVM against Soft Independent Modeling of Class Analogy (SIMCA), Partial Least Squares Discriminant Analysis (PLS-DA), Linear Discriminant Analysis (LDA), and Quadratic Discriminant Analysis (QDA).

Main Results:

• SVM modeling demonstrated excellent performance in both model transferability and large-scale classification.
• SVM, particularly with a hierarchical scheme, outperformed conventional chemometric techniques.
• MicroNIR spectrometers proved effective for NIR-based pharmaceutical RMID.

Conclusions:

• Portable MicroNIR spectrometers coupled with SVM modeling are highly effective for on-site and in-situ pharmaceutical RMID.
• This approach facilitates large-volume applications, enhancing pharmaceutical quality control.
• The study highlights the potential of integrated spectroscopic and chemometric solutions for the pharmaceutical industry.