Although the topic of tooth fractures has been extensively analyzed in the dental literature, there is still insufficient information about the potential effect of enamel microcracks (EMCs) on the underlying tooth structures. For a precise examination of the extent of the damage to the tooth structure in the area of EMCs, it is necessary to carry out their volumetric [(three-dimensional (3D)] evaluation. The aim of this study was to validate an X-ray micro-computed tomography ($$\mu $$
μ
CT) as a technique suitable for 3D non-destructive visualization and qualitative analysis of teeth EMCs of different severity. Extracted human maxillary premolars were examined using a $$\mu $$
μ
CT instrument ZEISS Xradia 520 Versa. In order to separate crack, dentin, and enamel volumes a Deep Learning (DL) algorithm, part of the Dragonfly’s segmentation toolkit, was utilized. For segmentation needs we implemented Dragonfly’s pre-built UNet neural network. The scanning technique which was used made it possible to recognize and detect not only EMCs that are visible on the outer surface but also those that are buried deep inside the tooth. The 3D visualization, combined with DL assisted segmentation, enabled the evaluation of the dynamics of an EMC and precise examination of its position with respect to the dentin-enamel junction.