Finding the needle in the haystack

Many conventional analytical techniques provide averaged data for the whole sample, but what if this doesn’t produce the information required? What if product performance depends on the physical location of a substance? Linda Kidder says that near infrared chemical imaging could be just the thing to find the needle your haystack.

The development of improved imaging technology and the computing power necessary to combine it with well-established analytical techniques is bringing new methods to the industrial analysis arena. Near infrared chemical imaging (NIR-CI), which as the name suggests combines conventional NIR spectroscopy with advanced optical imaging, is a prime example. Here we look at how this method works, the data that can be generated and its practical relevance.

With standard NIR spectroscopy a single averaged spectrum is produced for each sample. The spectrum shows how light in the wavelength 700 to 2500 nm has been absorbed, information that correlates directly with the presence of certain chemical bonds, primarily O-H, N-H and C-H. As these bonds are the key constituents of many organic molecules the technique is widely used for compositional analysis.

NIR-CI produces not one, but tens of thousands of spectra, generating spatially resolved data for a grid of individual areas or pixels, across the surface of a sample. A typical analysis will capture around 80,000 spectra over a period of several minutes. This amount of data demands detailed statistical analysis for effective interpretation, and a range of mathematical methods are applied. The underlying theories are complex but the best software packages can deliver the required chemometric and morphological tools in a user-friendly format. As a result it is possible to identify not only the specific substances present in the sample but also their location.

The information required from this analytical technique differs from application to application and data processing is tailored to reflect this. For example, in the development of engineered tablets, designed to deliver a specific drug release profile, the location of active ingredient in different parts of the tablet structure may be crucial. Alternatively the assessment of component abundance or blend homogeneity may be the more modest goal. The following examples highlight different types of application for which NIR-CI is particularly suitable.

Key features of NIR-CI analysers The principal components of a NIR-CI analyser are - an appropriate radiation source, a tunable wavelength filter and perhaps most importantly a two-dimensional array detector. Quartz lamps are a safe and easy to use option to provide NIR radiation, a liquid crystal tunable filter enabling wavelength discrimination, while simultaneously  maintaining image quality. Refractive optics and diffuse reflectance techniques are used, allowing analysis of most solid samples with no preparation. This is an important factor with respect to overall analysis time and the industrial relevance of the technique. The use of two-dimensional array detectors distinguishes a true NIR-CI analyser from a more basic mapping instrument which employs either a line array or single detector and has a step and acquire mode of operation. With these instruments the sample is moved relative to the detector during analysis, a requirement eliminated when using two-dimensional arrays. A NIR-CI analyser therefore has no moving parts, increasing its suitability for at- or near-line applications. Simple optics with long working distances is also important. This allows the analyst to configure either a small field of view – a single granule or tablet, or a larger area – a complete blister pack, for example, depending on the application. Flexibility is therefore an integral part of the design.

One of the challenges faced by pathologists is the detection of contaminants in tissue samples; medical silicone is one of many possible examples. Silicone was widely used for both cosmetic and reconstructive breast surgery prior to a ban in the US in 1992 because of possible links between silicone and autoimmune disease; implants were known to leak and/or rupture depositing silicone in the body. This study highlights the use of NIR-CI for the detection of silicone in human tissue.

Samples of tissue were imaged using two different fields of view - 3.2  2.6 mm (10 m/pixel) and 12.8  10.2 mm (40 m/pixel). The first level of magnification provided detailed information for method development; the second gave more rapid screening. Figure 1 shows images created by an initial mathematical analysis designed to highlight areas of dissimilar composition. The images correspond to about 75% of the original field of view. The first (Figure 1a) shows small areas of one component (red) surrounded by larger areas of a different material (green), which appears to make up the bulk of the sample. Closer analysis of the larger area allows it to be sub-divided into two different regions shaded blue and green (Figure 1b).

Figures 1a and 1b: Image analysis of a piece of silicone contaminated human tissue.

This initial analysis provides insight into the heterogeneity of the sample - the next step is to determine the chemical species that are present in each area. The interrogation of single spectra for each pixel allows the red material to be identified as silicone, while the green and blue areas are identified as proteins of slightly different type. From these single spectra, unique spectral features - marker bands - that differentiate the materials from one another can also be determined.

Using the marker bands, in combination with threshold values, each pixel can either be classified as one of the components or a mixture, or it can be unassigned. A red-green-blue image is produced, in which red, green and blue pixels represent the three components, and unassigned pixels are black (see figure 2). Unassigned pixels result from unidentified species or holes/thin patches in the surface.

Figure 2: A ’red-green blue image’ of a piece of silicone contaminated human tissue

Imaging pure samples of components possibly present in a sample also provides useful supporting data for the analysis. In this case pure samples of silicone and collagen, types I and III, were imaged, the resulting data confirming the red areas as silicone, while allowing the blue and green areas to be tentatively ascribed to type I and III collagen respectively.

Figure 3 shows the end result of the development work - a complete piece of tissue is rapidly screened allowing the efficient detection of localised areas of silicone contamination1.

For some applications, this one being an example, other analytical methods such as mid-infrared (MIR) and Raman imaging would provide a greater degree of chemical specificity. As these results show, however, NIR-CI offers sufficient chemical specificity coupled with fast analysis times and the capability to screen large areas. It is often therefore the best option for applications involving the detection of a contaminant in a relatively large area.

Figure 3: Imaging of a complete sample of silicone contaminated human tissue

Blend homogeneity is an important issue for the pharmaceutical industry impacting directly on product quality and dose uniformity. When assessing blend homogeneity, averaged data for the sample may not provide the required information at the micro level, so conventional spectroscopic and chromatographic techniques can be of limited use. For tablets, dissolution tests may highlight discrepancies in drug release behaviour, which may be linked with heterogeneity in the tablet, but will not yield information about the causes. NIR-CI, on the other hand, provides a non-destructive method for investigating both tablets and powders, as the following example demonstrates.

Six different tablets were analysed to assess the impact of blending time on homogeneity. Five of the tablets were produced specifically for the study by blending for different lengths of time; the sixth is a commercially available analogue. All of the tablets contain 80mg Furosemide, the active ingredient (API), and 240mg of an excipient mix. The time taken to capture an image of each tablet was around 2 minutes.

Figure 4 shows the images produced - the blue area is excipient and the green API. Sample F, the commercially available tablet, is clearly the most homogeneous, but E, the best-blended experimental tablet, also exhibits good uniformity with only a small area of unblended API. It is clear that as blending time is increased, from A through to E, the degree of homogeneity improves, as would be expected.

Figure 4: Images of six tablets produced from powders blended for increasing periods of time

Because of the quantities of data produced by NIR-CI is qualitative, visual assessment can be statistically investigated. An analysis of the distribution of intensities seen in the image (see figure 5) shows that a narrow Gaussian distribution is indicative of a homogeneous sample (type F, for example) while poor blending is associated with a broader peak shape. With tablet A, the least well-blended sample, two distinct populations, the unblended API and the excipient, are clearly visible.

For this application then the relative width (%STD) of the API, determined from the histogram of the chemical image, is a highly sensitive parameter for quantifying blend homogeneity. Development of a fully automated method for blend assessment could therefore be based on this variable2.
 
Advances in NIR-CI systems have brought them to a performance-price point that makes the technology valuable for a range of industrial applications. In some cases it is the ability of NIR-CI to generate spatially resolved composition data that is critical. In others it is the chemical specificity of the technique coupled with the ability to rapidly screen relatively large areas that is key. In almost all instances the ability to non-destructively analyse samples without any preparation is an important advantage.

Figure 5: Histograms showing intensity distributions for six tablets

Examples such as those included here illustrate the practical relevance of NIR-CI and how it can be used as a problem-solving tool. Industrial use of the technique, across a range of sectors is expected to grow as this relevance becomes increasingly well understood.

References:
1. Lee, E., Kidder, L.H., Kalasinsky, V.F., Schoppelrei, J.W. & Lewis, E.N. Forensic Visualization of Foreign Matter in Human Tissue by Near Infrared Spectral Imaging: Methodology and Data Mining Strategies, Cytometry, 2006, PtA, 69A, 888-896

2. Lyon, R.C.; Lester, D.S.; Lewis, E.N.; Lee, E.; Yu, L.X.; Jefferson, E.H.; Hussain, A.S., Near-Infrared Spectral Imaging for Quality Assurance of Pharmaceutical Products: Analysis of Tablets to Assess Powder Blend Homogeneity; AAPS PharmSciTech., 2002, 3(3), article 17, 1-15

Dr Linda Kidder is product manager chemical imaging systems. He graduated from Williams College with a B.A. in Chemistry, and received her Ph.D. in physical chemistry from the Johns Hopkins University where she was the recipient of the Sonneborn and Ernest Marks Fellowships.