Scaling pre-clinical drug discovery and development to meet high-volume manufacturing requirements is a challenge for pharmaceutical manufacturers. Spectral imaging and, in particular, hyperspectral imaging represent an opportunity to gain greater analytical understanding and improved process control capability with the deployment of "in-line" or "at-line" hyperspectral instruments.
Spectral imaging encompasses two techniques. One, Raman imaging, measures the molecular vibrational shift associated with exciting a sample compound with a specific laser frequency. The other technique that is emerging, hyperspectral imaging, is a highly specific analytical approach for pharmaceutical manufacturers that some say has advantages over Raman imaging, including:
• It has a large field of view for wide area sampling versus single tablet analysis;
• It has high-throughput screening for real-time analysis and control;
• When utilized in the near infrared (NIR) or short wave infrared (SWIR), hyperspectral instruments offer much greater spectral specificity than mid-wave infrared (MWIR) techniques;
• No sample preparation or sample handling are required; and
• Sampling is in line and non-destructive.
Consistent with the goals of the Food and Drug Administration’s (FDA) process analytical technology (PAT) initiative, hyperspectral imaging instruments greatly enhance the knowledge and understanding of the pharmaceutical process by capturing all of the spectral and spatial attributes of material samples within the sensor’s field of view (FOV).
When combined with spectral libraries established during the drug discovery phase and multivariate analytical models, hyperspectral sensors such as Headwall Photonics’ Hyperspec NIR (900 to 1700 nm) or the Hyperspec SWIR (1000 to 2500 nm) can make accept or reject decisions when deployed in line or at line. As a result, hyperspectral imaging allows pharmaceutical manufacturers to establish critical control points (CCPs) from the post-discovery phase through pre-production and high-volume manufacturing.
Applying Hyperspectral Imaging
Utilizing high-efficiency diffractive optics, hyperspectral sensors can be configured to offer peak optical efficiency in wavelengths across broad spectral regions. With the underlying advantage being deployment for wide area spectral analysis over a conveyor processing line versus an off line, tablet-by-tablet approach, hyperspectral sensors enable a new set of application capabilities as pharmaceutical manufacturers ramp to volume with precise control over key steps in the production process.
Pharmaceutical manufacturers can use hyperspectral analysis to increase production yields in several areas, including active pharmaceutical ingredient (API), content uniformity, polymorph analysis, quality control over spray dry dispersion, and anti-counterfeit verification and authentication.
In a post-discovery production application, capturing precise spectral information from pharmaceutical manufacturing control points has traditionally involved either simple machine vision or single-point NIR spectral instruments deployed off line. The limitation of these systems is that they are only capable of sampling a very small area of the overall product flow and do not lend themselves to high-speed production processes. In addition, these options are costly due to redundant equipment, poor sampling rates, and time required for analysis.
Defining Hyperspectral Analysis
Hyperspectral imaging technology is an established chemical sensing and imaging technology that allows for the spectroscopic analysis of any particular sample or point within a scene of interest. After its introduction during the mid-1980s as a means to conduct remote sensing experiments and its success as a military and defense sensor technology, hyperspectral imaging is poised for rapid adoption in pharmaceutical applications involving the complex manufacturing of chemical materials and products.
Within the post-discovery manufacturing environment through the ramp to high-volume manufacturing, hyperspectral imagers capture and build a wavelength intensity map of a scene with high spatial resolution. The combination of spectral data and spatial detail enables the high-speed analysis of chemical content, uniformity, quality, and a host of other spectral characteristics and attributes. Traditionally, because these hyperspectral imaging systems were designed to perform under ambient lighting conditions such as available sunlight, they required innovative instrument designs to optimize environmental parameters such as signal-to-noise, optical efficiency, and dynamic range.
Hyperspectral imaging yields the following results within a pharmaceutical manufacturing operation:
• A rendered view of the scene of interest based on known chemical spectra or established spectra libraries;
• For in-line or at-line deployment, spectral wavelengths of interest can be interrogated based on defined intensity thresholds as material and samples pass by the hyperspectral imaging sensor; and
• For any point or pixel within the FOV, the chemical spectra or spectral signature of any particular point can be determined while maintaining the integrity of spatial information obtained.
All hyperspectral imaging instruments consist of the following key elements: a high-performance, aberration-corrected imaging spectrometer; a fore-optics lens selected for the appropriate field of view and distance from the production line; stable illumination; a reference tile such as Spectralon (a reflectance material); a processing unit or computer; and application software for acquiring the images, creating the hyperspectral data cube, and rendering a control decision. Often, depending on the harsh nature of the production environment, the instrument enclosures may be rated for industrial use. For pharmaceutical manufacturing operations, transmissive optics or prisms are not used within the spectrometer, because high imaging performance is required and these components lead to excessive stray light within the system.
Hyperspectral imagers are deployed as a scanning "push-broom" spectral imager. For each moment in time or frame capture by the sensor, the scene observed by the instrument fore-optic lens is imaged onto a tall slit aperture of the hyperspectral instrument. The scene that fills the slit aperture of the sensor is re-imaged through the spectrometer with the wavelengths dispersed by a diffractive grating onto a two-dimensional focal plane array (FPA) such as a charge coupled device.
One axis of the focal plane array (pixel-rows) corresponds to the imaged spatial positions within the FOV all along the slit height. The second axis (spectral for pixel-columns) corresponds to the spectral wavelength that is linearly dispersed and calibrated. Each two-dimensional image or frame capture is digitized by the FPA to build a dataset that comprises all of the spectral and spatial information within the scene or FOV of the sensor.
While scanning a wide conveyor of moving tablets or pills, multiple two-dimensional image frame captures are rapidly taken as tablets pass by the hyperspectral imager; these individual frames are taken at very high speed and are stacked like a deck of cards to produce a data file commonly called a hyperspectral data cube. The value of each pixel within this hyperspectral data cube represents the wavelength-calibrated spectral intensity of that pixel’s small FOV on the scene.
Imaging performance attributes that are critical to the successful deployment of hyperspectral sensing for pre-clinical development in pharmaceutical manufacturing operations are the achievement of high spectral and spatial resolution and exceptional photometric accuracy.
To demonstrate the value of hyperspectral imaging technology, Headwall application engineers scanned four tablets of similar appearance and size, composed of the following generic drug compounds: aspirin, acetaminophen, vitamin C, and vitamin D. These four compounds were scanned simultaneously with NIR hyperspectral imaging sensors utilizing a moving linear stage to simulate a conveyor line manufacturing process. The samples were scanned in both powder and tablet form and the corresponding image slices were "stitched" together.
One obvious advantage of hyperspectral imaging is the ability to scan multiple tablets simultaneously as they move across a process line. The number of tablets that can be scanned is based on required FOV and the spectral and spatial resolution required (instantaneous field of view). These parameters are application specific and can be modeled to identify the specific sensor components needed to achieve the required performance. Once key spectral features are identified, the hyperspectral imager can be set to bin areas of pixels or regions of interest in order to achieve high volume throughput.
The diagram in Figure 1 (see p. 36) represents a single image slice of the four generic tablets, taken during a process line scan experiment. The corresponding "blur" scan in the photograph represents a magnified slice of the FOV as the tablets move across the process line; the data content comprises all of the spectral and spatial information that will used to build the hyperspectral data cube from which spectral imaging data will be rendered and analyzed.
Hyperspectral sensor technology analytical information can be displayed in real time while the entire analytical data set (spectral and spatial) is captured and stored for further analysis at a later time. Applications involving NIR chemical imaging such as spray dry dispersion, content uniformity, or polymorph analysis are well suited to hyperspectral imaging techniques. Standard applications like bulk quality control, blending, and packaging also benefit from hyperspectral imaging.
For any pixel or spatial position within the FOV, the corresponding chemical spectrum is obtained and can be graphed as required. Figure 2 (see p. 37), for example, shows the spectral imaging results generated as a result of the hyperspectral scan of the generic tablet compounds. For the purposes of this illustration, all of the corresponding spectra for each of the pharmaceutical tablets are shown.
Can Be Customized for CCP
Given the specific pharmaceutical application or analytical need at that CCP, specific or relevant spectral data can be rendered in the required level of detail or wavelength resolution based on the speed of the production environment. For example, not all hyperspectral data need to be displayed; in some cases, it may be appropriate only to display spectral bands of interest to the researcher or manufacturing engineer.
Compounds and chemical formulations comprised of different spectral signatures can be color coded based on user-defined parameters and established spectral libraries developed during the drug discovery phase. For example, with spray dry dispersion techniques, the presence of multiple compounds or lack of spectral homogeneity in a sample or pill of interest can be identified as in the hyperspectral image in Figure 3 (above, left).
The power of the hyperspectral NIR imaging technique becomes apparent when standard techniques of spectroscopy are applied to the spectral signatures of each tablet. The results of applying hyperspectral spectroscopy to pharmaceutical tablets in the post-discovery phase show that high-speed screening extracts sufficient information from the spectral imaging data to allow discrimination of each tablet from the next with complete certainty.
In demonstrating the spectral features derived from hyperspectral scanning, Figure 4 (see p. 38) displays a series of derivatives taken of the spectra for a series of five random spatial points of each tablet. By applying the standard methods of spectral chemometric analysis and creating a special material index to define each tablet as an independent material variable, the chemometric model can recreate with a high degree of statistical certainty the material index of that particular table or of a similar tablet.
These techniques can be applied to the movement of tablets in real time, and early work suggests that tablet discrimination can be achieved at a very high throughput rate. In many cases, the technique can also be applied to the real-time quantification of the material makeup of tablet samples within the same material population (see Figure 4).
While hyperspectral imaging has been established as a proven, hardened technology for the harsh environments of military, defense, and remote sensing deployments, the technique’s use in pharmaceutical manufacturing operations has demonstrated considerable value in the past few years. Understandably, a key driver in the adoption of hyperspectral imaging instruments is the FDA’s PAT initiative to gain increased control and process understanding at various points in the production process.With the introduction of commercially available hyperspectral instruments, these spectral imaging sensors can be deployed to increase production yields in a cost-effective manner at many points along the production process, with attractive return on investment and very short payback periods