(M1530-11-60) Impact of X-Ray Micro-CT Resolution on Quantification of Quality Attributes and Simulated Mass Transport Properties of Lyophilized Products

Purpose: A fine balance between resolution and field-of-view is needed when using an image-based approach to assess the microstructure of lyophilized products. Typically, higher resolution images reveal in-depth microstructural features yet have limited fields-of-view. The limited field of view may lead to inaccurate assessment of imaging data and quantification of product attributes. Alternatively, low resolution imaging yield images representative of the sample but may not accurately capture all microstructural features that impact process efficiency and product performance. Thus, it is key to select the optimal resolution that resolves the features of interest while also ensuring representativeness. The goal of this study was to systematically assess the impact of imaging resolution on the quantification of quality attributes of lyophilized cakes. In this study, X-ray microscopy (XRM) was utilized because of its high resolution, non-destructive, and three-dimensional capabilities.

Methods: The sample was frozen at -45 ^oC before sublimating at -18 ^oC and 300 mTorr in a Millrock MicroFD freeze dryer. The cake was subsequently scanned seven times with a Zeiss Versa 520 system at multiple resolutions of 1, 2, 5, 10, and 22 μm/pixel. All datasets were analyzed using digiM I2S software that utilized artificial intelligence (AI)-based approaches to rapidly and accurately assess the impact of resolution on subsequent analysis. Machine learning algorithms were used to segment the pore and solid phases of the lyophilized cake. From this segmentation, the overall cake porosity, pore size distribution, and cake wall thickness distribution were determined. Simulation of mass transport properties were conducted through the solving of computation fluid dynamic equations based on the digitized cake structures previously quantified.

Results: With XRM, the interior microstructure of the cake can be imaged non-invasively within the glass vial, thus, not subjecting the sensitive cake to humidity damage. The 22 μm/pixel resolution XRM scan was able to capture the full cake morphology, as shown in Figure 1. The cake displayed structural heterogeneity; the top 20% of the cake showed smaller pores and thinner walls than the bottom 80%. Higher resolution scans were conducted throughout the cake height capturing (Figure 1) the heterogeneous microstructural differences. As resolution decreased, the thinner solid walls became impossible to resolve. After AI-based segmentation, the cake microstructure was quantified, specifically the fraction of pores and solids, pore size distribution and cake wall thickness. The volume fractions of porosity ranged from 88.5 to 94% and decreased as resolution decreased as seen in Figure 2a. Conversely, the average pore size and cake solid wall increased as resolution decreased. The average pore size average changed from 160 to 1400 μm while the cake solid wall thickness changed from 3 to 63 μm as resolution decreased (Figure 2b, 2c). The volume fractions did not significantly change with resolution, but the pore sizes and thicknesses of cake solid walls were significantly affected. The spatial distributions of the solid phase showed variation by 2% in the axial and radial directions regardless of resolution (Figure 3a).  Permeability values calculated by simulating water vapor transport through the digitized microstructure generally increased when resolution decreased as shown in Figure 3b.  2 μm/pixel was determined to be the optimal resolution to capture a representative field-of-view while still retaining high enough resolution to resolve the pores and cake solid walls accurately.

Conclusion: In this study, the structural characterization of lyophilized cakes via quantitative X-ray microscopy was critically assessed to determine the optimal imaging conditions for representativeness and analysis of the features of interest. As resolution decreased, porosity decreased slightly while pore size and cake solid wall thickness greatly increased. Radial and axial spatial distributions showed slight variations from average solid volume content for all resolutions. Permeability values increased overall as resolution decreased. By quantifying the microstructures with AI and simulating transport properties to assess the effect of resolution, 2 µm/pixel was determined to be the optimal resolution for imaging lyophilized cakes. Identification of the optimal resolution will allow for targeted structural characterization more rapidly. This can be applied to investigate advanced lyophilization methods like controlled ice nucleation for optimized cake structure to improve primary drying times.

Figure 1: A representative slice from each X-ray microscopy scan at different resolutions on the same lyophilized cake sample showing the microstructure. The preview from the full cake scan shows the location of where each high resolution scan was conducted. The cake solid is the bright phase and the pores are the dark gray phase in the scans.

Figure 2: Correlation plots with resolution of the image data volume on x-axis and a quantitative measurement of the microstructure on the y-axis: a) porosity volume fraction, b) average pore size, and c) average cake solid wall thickness. An exponential best-fit line is applied to the data.

Figure 3: a) Axial spatial distribution plot with the average solid volume fraction % on the y-axis and distance from bottom of the image dataset on the x-axis. Inset on the bottom right is a schematic of how the distribution is calculated. b) Radial spatial distribution plot with the average solid volume fraction % on the y-axis and radius from center of the image dataset on the x-axis. Inset on the bottom right is a schematic of how the distribution is calculated. c) Permeability curves through simulation of the imaged dataset.