As published in Manufacturing Chemist, Jan 2012
By Dr. Brady Carter
Keywords: Quality Control, PAT, QbD, Counterfeits, Imaging
Multispectral imaging unlocks valuable information on both quality and safety of pharmaceutical materials. Parameters measureable by multispectral imaging such as colour, texture, gloss, shape, and size provide useful information on a finished product, on the raw materials, or on the manufacturing process itself. Multispectral imaging provides a rapid quality assessment of uniform and inhomogeneous samples alike by combining information on all of the above parameters automatically in just a few seconds, requiring no sample preparation and leaving the material intact.
Figure 1: VideometerLab4: Up to 20 wavelengths via LEDs, uniformly spaced around the inside of the integrating sphere are strobed successively each generating a monochrome image.
The earliest applications for multispectral imaging focussed in the food industry as replacements for subjective and expensive sensory panels. As these applications have become widespread, other highly regulated industries, such as pharmaceuticals, have begun to reap the benefits from multispectral imaging in ensuring high quality products, and in the battle against counterfeiters.
Traditional colour imaging uses three broad bands of colour red, green and blue and is known as RGB imaging. As a consequence of the broad bands RGB imaging has very limited spectral resolution and is unsuited to differentiating samples showing variation within a single broad band. Multispectral imaging refers to multiple wavelengths over the whole range from UV through visible to NIR (230 to 1050 nm). VideometerLab 4, a lab based multispectral imager from the Danish based company Videometer A/S, is based on a high-intensity integrating sphere illumination featuring light emitting diodes (LED) and a black and white high resolution CCD camera (2056 x 2056 pixels) (Figures 1 & 2). Measurements are combined at up to 20 different wavelengths into a single high resolution multispectral (2056 x 2056 x 20) image with every pixel in the image representing a spectrum.
Figure 2: Schematic of integrating sphere: Internal reflection of the LEDs by the diffused white inner surface of the sphere ensures diffused homogeneous light for increased reproducibility, dynamic range, low scatter, and shadow effects.
The non-destructive investigation of materials with non-uniform colour and texture can be difficult, tedious, and expensive. Conventional techniques such as NIR spectroscopy only measure a single point or average over a fixed area, and do not give an objective overall assessment of visual quality. Multispectral imaging can be described as a trade-off, sacrificing some spectral resolution to increase spatial information giving a ‘snapshot’ of the combined bulk properties of a sample, handling natural variation and inhomogeneity.
The choice of illumination wavelength ensures each application can be optimised and is not restricted to the wavelengths spanned by traditional RGB technology. Using LEDs in the UV or NIR adds information not visible to the human eye. As an example, most objects are white or transparent in the NIR region, which allows for the separation of the colour and surface properties of the measured object. Uniform and non-uniform samples alike are simply placed in the target area and a custom designed PC software for data capture and analysis means results are available in less than 10 seconds; including sample handling time. A radiometric and geometric calibration procedure with NIST traceable standards is available to ensure accuracy and repeatability, and automatic diagnostic tests can be performed routinely to ensure instrument stability.
QC/ QA Inspection
Figure 3: RGB and pseudo images showing the identification of increased active ingredient within the formulation of a single tablet.
Multispectral imaging offers a rapid solution to screening for out of specification tablets and contaminated pharmaceutical powders. Since the response of the test material is measured simultaneously at up to 20 different wavelengths, very subtle changes in composition can be detected and imperfections are automatically flagged. This test is entirely non-destructive requiring only a sub-sample of the product to be placed in a petri dish before analysis is carried out.
The technique is perfectly suited for the objective measurement of tablet ‘Elegance’, a standard pharmaceutical test in which minor, major, and critical defects are determined on a sub-sample; leading to acceptance or rejection of an entire batch. Tablet features inspected using multispectral imaging, both in the laboratory and on-line, include:
- Shape integrity (no missing parts)
- Tablet color
- Foreign material (impurities, spots)
- Cracks, picking and chipping
- Dispenser colour and defects
Figure 4: RGB and pseudo images showing automatic discrimination between coated and uncoated mini-tabs that are visually identical. This allows the operator to accurately quantify the level and homogeneity of tablet coating.
Understanding how well the ingredients of pharmaceutical formulations are blended is critical across various stages of the production process. Multispectral imaging can be used to recognise different powders in single layer and bulk presentation. As a result, the success or otherwise of bulk mixing of two powders can be quantified.
Figure 5: RGB and MNF (Maximum Noise Function) transformed images of pharmaceutical powder highlighting dirt and dust contamination. MNF is the imaging equivalent of Principle Component Analysis (PCA).
Figure 6: Each component within the mixing process is assigned a weighted index based on its spectral profile. The ‘Mix’ value indicates to the operator that the make-up of the sample is from ‘Powder A’ and ‘Powder B’.
The presence of counterfeit pharmaceutical products poses an obvious hazard to human health. Multispectral imaging can be used to identify counterfeits through analysis of both packaging and product. Although visually it may be difficult to detect counterfeit packaging, it is a simple matter to automatically compare the shape and size of characters, trademarks, company names and logos on good/bad packaging. The use of specialised inks and texture effects on authentic packaging is simple to detect.
Figure 7. A comparison between authentic and counterfeit packaging with multi-spectral imaging highlighting differences seen by multispectral imaging that would not be obvious through visual inspection alone.
Characterizing the product itself is also straightforward. Subtle changes in size, shape, colour, and markings can all be combined and used to automatically detect counterfeits. Once presented with examples of counterfeit products, the VideomaterLab 4 can learn to identify counterfeit products when presented with unknown samples. Alternatively, it can simply look for differences among a sample of unknown drug products and identify any products that vary, although without knowing which products are legitimate and which are counterfeit.
Figure 8: RGB and pseudo images of authentic (right) and counterfeit (left) drug.
The VideometerLab 4 is even able to identify non-conforming product in the package without removing the product from the original blister packaging or bottle (Figures 8)
Figure 9. The authentic tablets on the right can be differentiated from the counterfeit tablets on the left using multi-spectral imaging while still in the blister pack.
Multispectral imaging technology has the potential to become an important tool for measurements of non-homogeneous samples in the pharmaceutical industry. High-performing and relatively inexpensive systems are available to provide accurate results in a broad range of critical applications.