Category: Formulation and Quality
Purpose: A stable and robust freeze-drying process is of high desire in pharmaceutical lyophilization to ensure homogenous and intact cake appearance. Process parameters and physio-chemical properties, such as freeze-dried cake moisture content, reconstitution time, integrity, homogeneity, uniformity and potency, are often employed to confirm drug product quality as well as its stability under storage. Despite decades of collective experience, lyophilization process remains to be time-consuming and costly in both development and scale-up phases due to the challenges in cycle development, process optimization, and cake characterization. In this study, we explored the application of a high-resolution, non-invasive imaging with X-Ray Microscopy (XRM) to collect three-dimensional (3D) volume data from lyophilized samples non-destructively, followed by quantification with an artificial intelligence (AI)-based image processing method as a new powerful tool for quantitative and qualitative cake structure characterization.
Methods: In this study, an XRM system with a spatial resolution of 1µm was applied to scan a lyo cake sample with and without glass vial in three dimensions during the imaging experiments. The X-ray was able to penetrate through the glass vial and collect images of the internal sample noninvasively. The 2D cross section images from the 3D volume was shown in Figure 1. The massive imaging data collected during the high-resolution 3D scans were processed with the cloud computing interface DigiM I2S. With AI-based image segmentation tool, important physio-chemical properties of the lyo cake were computed and the 3D structure of the cake was reconstructed.
Results: The lyo cake sample was scanned using XRM at resolution of 1mm without being taken out from the glass vial first and then scanned again after being removed from the vial. Figure 1 compares the 2D cross sections of two XRM scans of cake sample, where the cake microstructures within (Figures 1a and 1c) and outside (Figures 1b and 1d) glass vial. Figures 1a and 1c showed a successful cake-in-vial scan without any beam hardening artifact. Porosity and thin wall of protein solid were both resolved clearly. Figures 1b and 1d showed a scan of the same cake sample taken out of its container. Small deformation of the protein solid has been captured during the scan as double-edged features, denoted by the red arrows in Figure 1d. The comparison clearly indicated that the sample underwent some physical and maybe chemical change while taken out from the vial, thus undesirable. The subtle change of the protein solid layers was captured and was correlated with the mechanisms of cake collapses, thanks to the high resolution and contrast of XRM.
The AI-based software, DigiM I2S was used to segment the images and reconstruct the 3D structure of the lyo cake. A snapshot of DigiM I2S’ AI-based image segmentation interface was showed in Figure 3a. The traces of lines were defined as the seed to train DigiM I2S AI engine. Red trace was the training seed for the porosity, and green trace for protein solid. The image was classified into two phases, red phase for porosity and green phase for protein solid (Figure 3b). The cloud computing software then batch processed the whole deck of images using user defined and accepted training set (Figure 3c). Last but not least, the 3D cake structure was reconstructed with the full volume image data, as shown in Figure 2d.
Conclusion: Image-based characterization has great potential in a number of lyophilized drug product characterization. XRM is able to qualitatively and quantitatively assess cake integrity and purity non-invasively, which is not achievable by other techniques. In addition, parameters such as porosity that are challenging to be determined experimentally, can be assessed using image data. This approach can be used to assess inter-batch and intra-batch homogeneity as well as cake stability upon storage with limited amount of sample. The physio-chemical properties obtained can be used for fundamental understanding and correlation with process variation and optimization. On the other hand, XRM also presents limitations which should not be overlooked. The finite resolution limits the smallest feature that can be studied. Balancing the representative sample size and optimal resolution is critical. With the image data progressively develops into a digital drug database, the management, visualization, quantification of simulation of massive amount of data requires dedicated hardware and software. Albeit these challenges, X-ray based imaging technologies such as XRM at present is rapidly evolving to provide meaningful cake characterization information, guiding formulation as well as process development throughout the lyophilized product development and manufacture.