On precise localization of boundaries between extended uniform objects in MRI: tooth imaging as an example

The purpose of this study was to investigate the achievable precision of localization of boundaries between extended uniform objects in MRI and to study the effect of zero-filling on reaching it.

A theoretical model of an object boundary in the presence of noise was introduced, and the error of localization was derived. The effect of zero-filling on reaching the achievable precision was assessed by computer simulations and experimentally on an extracted tooth in a signal-giving medium.

With the help of the theoretical model, the achievable precision of localization of boundaries between two uniform extended objects was shown to surpass the nominal resolution by a factor equal to the contrast-to-noise ratio. In the simulations and phantom experiments, zero-filling followed by image segmentation allowed for approaching the theoretical value. As an application example, an MRI-based dental impression was performed in vivo, and a bridge was produced and permanently fixed to the volunteer's teeth.

This work demonstrates that in an MRI experiment, the achievable precision of localization of object boundaries is not limited to the nominal resolution and can surpass it by an order of magnitude. Zero-filling is a simple and effective method of reaching it.