(301aa) Integrating Multi-Variate Image Analysis and Artificial Intelligence Techniques with Pvm for Inline Crystal Size and Shape Measurements

On-line monitoring of particle shape and size distribution is a challenge in the pharmaceutical and fine chemical industries due to process complexity, the lack of process understanding and adequate in-situ sensors. However, with the advances in real-time imaging hardware, such as video cameras and fiber optics (exemplified by Focused Beam Reflectance Measurement, FBRM and Particle Vision and Measurement, PVM, both from Lasentec) combined with regulatory initiatives such as the US Food and Drug Administration's (FDA) Process Analytical Technology (PAT) program, there is a growing interest in the pharmaceutical and chemical industries as well as the research community to develop advanced in-line control technologies for particulate processes using such advanced imaging sensors.

A number of studies (Yu et al. [1]; Birch et al. [2]; and Barrett et al. [3]) have recently highlighted the success of implementation of those imaging sensors to gain insights into crystallization process and thereby provide extra capability for in-line process control. However, there is still a gap between the information obtained from advanced imaging sensors and the knowledge required for in-line control of crystallization process. While variables such as crystal size and shape are critical in crystallization, available in-situ sensors do not provide direct measurements of these. Information about crystal size can be indirectly obtained in the form of chord length distribution as measured by FBRM, however in numerous situations, the chord length deviates systematically and significantly from the true crystal size distribution (Patience [4] and Calderon De Anda et al. [5]) In this paper, we aim to circumvent these challenges by using in-process video imaging (PVM in particular) for determining particle shape and size distribution.

Multivariate image analysis (MIA) uses multivariate statistical approaches to extract information from image data. MIA has been employed by MacGregor and coworkers for process monitoring and control [6-7]. In this work we combine Image Processing Techniques from artificial intelligence with MIA and for analyzing in-situ PVM images from a crystallization process case study. In our approach, Image Processing Techniques specifically image enhancement, edge detection, morphology operations, and feature extraction are employed to detect objects (crystals) of interest in the in-situ images. Each detected objects is enclosed within a minimum-area rectangle, from which morphological descriptors (e.g. shape factors) are derived. In parallel, multiway principal component analysis (PCA) is also performed on the composite image and statistical descriptors derived from the scores plot. The morphological and statistical descriptors are then jointly used to identify particle size and shape.

We have previously reported the reliability of this method for classifying a variety of particle shapes [8]. However, we notice that image segmentation is a critical and challenging task that determines the overall robustness of the method. A large variety of noise sources confound the image segmentation step including bubbles in the solution, out-of-focus objects, and motion of the particles. In this work, we describe advancements that improve the accuracy of determining the particle outline. Enhancements being investigated include image processing improvements through image enhancement, separating the object from the background by filtering and histogram equalization and combining information from multiple images to eliminate the effect of particulate motion. MIA techniques also allow localization of the particle location from the loading vectors. Image segmentation is then accomplished using statistical indices including Fourier statistics, covariance statistics, and other derived ones. In this paper, we describe the image enhancement techniques and their application to industrially important crystallization process. These results show large improvements in accuracy and robustness of the method, and yields dynamic real-time particle size distribution that can be used for closed-loop control.

References

[1] Yu L. X., R. A. Lionberger, A. S. Raw, R. D'Costa, H. Wu, and A. S. Hussain, Advanced Drug Delivery Reviews, 56 (2004) 349

[2] Birch M., S. J. Fussell, P. D. Higginson, N. McDowall, and I. Marziano, Org. Process Res. Dev., 9 (2005) 360. [3] Barrett P., B. Smith, J. Worlitschek, V. Bracken, B. O'Sullivan, and D. O'Grady, Org. Process Res. Dev., 9 (2005) 348

[4] Daniel Bruce Patience, PhD. Thesis (2002), University of Wisconsin-Madison

[5] Calderon De Anda J., X. Z. Wang, and K. J. Roberts, Chem. Eng. Sci., 60 (2005) 1053

[6] Yu H. and J. F. MacGregor, AIChE J., 50 (2004) 1474. [7] Yu H., J. F. MacGregor, G. Haarsma, and W. Bourg, Industrial Engineering and Chemistry Research, 42 (2003) 3036

[8] Y. Zhou, X. T. Doan, R. Srinivasan, Real-time Imaging and Product Quality Characterization for Control of Particulate Processes, ESCAPE 2006.