Image processing method

A method for training a machine learning inference model, wherein the trained machine learning inference model is for use in generating a representation of a virtual three-dimensional cast of an internal structure of an individual's intra-oral tissue from a computer-readable surface representation of exposed portions of the individual's intra-oral tissue, the method comprising:

• obtaining first three-dimensional image data representing tissue characteristics within an intra-oral cavity for a plurality of individuals;
• obtaining second three-dimensional image data representing a shape of a volume enclosed by an external surface of intra-oral tissue that is exposed within the intra-oral cavity for the plurality of individuals, wherein the second three-dimensional image data is of exposed intra-oral tissue that comprises both hard tissue and soft tissue, and wherein the first three-dimensional image data and second three-dimensional image data for each of the plurality of individuals form a co-registered pair;
• and
• training the machine learning inference model using a training set obtained from the co-registered pairs of first three-dimensional image data and second three-dimensional image data for the plurality of individuals, wherein the first three-dimensional image data is used as a target for the training.