Manufacturing & Analytical Characterization
Eric Gottlieb
Business Development
SentrySciences, Inc.
Longmont, Colorado
Description: Due to inaccuracies of light obscuration-based particle counting, particle imaging tools have long been used as an orthogonal method. The advent of AI tools has made simple qualitative classification of inherent particles possible, yet the approach has limitations and can introduce bias and risk. In addition, while AI classifiers are typically used with particle imaging to sort subvisible particles by type for size and counting purposes, at some point in the development process it is necessary to look at more than just size and count. This talk will discuss moving beyond the qualitative parts of classification to create fingerprints based upon subvisible particle images by using a convolutional neural network (CNN) to analyze the typical morphology of the particles. These fingerprints provide quantitative comparison data based upon morphological changes caused by stressors, process changes, or time.