Unmasking translucent protein particles by improved micro-flow imaging™ algorithms

Micro-flow imaging (MFI(™) ) is an increasingly important technique for the characterization of subvisible particles during the development of biopharmaceutical products. Protein particles are highly variable in size, appearance, and translucency posing challenges to optical detection techniques. We have developed a set of standard statistical tests applicable for routine evaluation of MFI™ particle dataset quality. The tests evaluate the spatial randomness of particles using nearest neighbor and quadrat analysis. Using case studies of stressed antibody samples, we demonstrate how the tests uncover fragmentation artifacts and uneven detector sensitivity for translucent particles. To improve the detection of translucent particles, a new local pixel intensity variance particle detection algorithm has been developed. The improved algorithm significantly decreases fragmentation artifacts, and also increases sensitivity toward translucent particles in general. Our results highlight current limitations and the potential for improvements in the optical detection techniques for subvisible protein aggregates.