Multidimensional particle characterisation from in-situ imaging using deep learning and transfer learning

Particle size and shape are important in the pharmaceutical industry, affecting both process efficiency and product performance. Quality-by-design and continuous manufacturing are aided with appropriate models of processes — selection and calibration of which are informed by measurement of particle size and shape. Off-line measurements have inherent limitations when following the trajectory of particle attributes in a process; removing and treating material for off-line analysis can alter particle characteristics. In contrast, in-line measurements provide representative measures of particle size and shape at the expense of producing more challenging (out of focus, overlapping particles) datasets for extraction of particle characteristics.