A new Process Analytical Technology (PAT) has been developed and tested for on-line process monitoring of a vacuum freeze-drying process. The sensor uses an infrared camera to obtain thermal images of the ongoing process and multivariate image analysis (MIA) to extract the information after automatic detection and segmentation of the region corresponding to the product in every vial. A reference model was built, using the information of six batches, and different kind of anomalous events, involving either single vials in the batch or the whole batch, were simulated to test the capacity of the MIA-based monitoring system to promptly identify them.