A method of correction of a three-dimensional digital model of an arch of a patient or “reference model.” The method includes the following steps: acquisition of at least one two-dimensional image of the arch, called updated image (Ia1;Ia2), in actual acquisition conditions; search, for each updated image, of virtual acquisition conditions providing a two-dimensional image of the reference model, which maximizes the fit with the updated image, or “reference image” (Ir1;Ir2); comparison of the reference image with the updated image to detect differences; and correction of the reference model to reduce said differences to get a corrected reference model.