High-Speed\nImaging-Based Particle Attribute Analysis\nof Spray-Dried Amorphous Solid Dispersions Using a Convolution Neural\nNetwork

Spray drying is a well-established\nmethod for preparing\namorphous\nsolid dispersion (ASD) formulations to improve the oral bioavailability\nof poorly soluble drugs. In addition to the characterization of the\namorphous phase, particle attributes of spray-dried intermediates\n(SDIs), including particle size, morphology, and microstructure, need\nto be carefully studied and controlled for optimizing drug product\nperformance. Although recent developments in microscopy technology\nhave enabled the analysis of morphological attributes for individual\nSDI particles, a high-throughput method is highly desirable. In this\nwork, a fingerprinting method exploiting high-speed dynamic imaging,\nlaser diffraction (LD), and a convolutional neural network (CNN) was\ndeveloped to characterize and quantify size and morphological distributions\nof particles in batches of spray-dried ASDs. This imaging technology\nenables the generation of hundreds of thousands of single-particle\nimages in a few minutes that are analyzed by both unsupervised and\nsupervised CNN models. The unsupervised data mining analysis demonstrated\nthat a batch of SDI is a mixture of diverse particle subpopulations\nwith varying sizes and morphological attributes. Motivated by this\nobservation, we developed a CNN model that enabled rapid computation\nof the volumetric composition of the distinct particle subpopulations\nin a SDI batch, thus generating a morphological fingerprint. We implemented\nthis high-speed imaging-based particle attribute analysis method to\ninvestigate SDIs containing hypromellose acetate succinate as a model\nsystem. The CNN fingerprint results enabled quantification of the\nchanges in the morphological distribution of SDI batches prepared\nwith variations in the spray drying process parameters, and the results\nwere in line with the LD and electron microscopy data. Our experiments\nand analysis demonstrate the robustness and throughput of this fingerprinting\napproach for quantifying particle size and morphological distributions\nof individual SDI batches, which can help guide spray drying process\ndevelopment and thereby enable the development of a drug product with\nmore robust process and optimized performance.