Comprehensive dental image analysis

In the dental field, image data (X-ray images) represent the initial source of information for a first assessment of a patient's health status and also serve as the basis for further planning of the treatment process. Both extraoral images, such as the orthopantomogram (OPG) or lateral cephalometric radiographs (LCR), and intraoral images, such as bitewing radiographs, are used. The OPG is the typical initial image because all teeth including roots are clearly visible.

More comprehensive analyses take time

The features recognizable in the images (caries, fillings, etc.) of the teeth must first be visually analyzed and then documented by the doctor. More comprehensive analysis methods, such as determining angles and relationships between manually identified landmarks in lateral cephalometric radiographs, require significantly more time and effort.

Automatically extracted information supports medical staff

The Visual Healthcare Technologies department is working on solutions that enable automatic analysis of image data and provide the extracted information to the doctor. As a fundamental step, the contours of the individual teeth are automatically found in the images. In this process, the domain-specific knowledge of experts is represented through a statistical model, which depicts both the relative positional relationships and the shape variance of the teeth. This model is initially placed on an image using deep learning. Subsequently, the teeth in the image can be identified and their contours found.

The existing orthopantomogram, a two-dimensional X-ray image of the upper and lower jaws.
The final segmentations of the individual teeth can now be used to find individual characteristics of these teeth and to provide them collectively.

Image-based 3D reconstruction of teeth

The department is also working on solutions to extract information from image data that is normally used only for documentation purposes. In the orthodontic treatment of dental misalignments, five photographs of the dentition are routinely taken, whose automatic evaluation can support the decision-making process and the monitoring of the treatment course. For this purpose, the tooth contours and numbering are first extracted in the individual photographs. Subsequently, the silhouettes generated in this way, which represent the teeth from different angles, are used to determine the surface and spatial arrangement of the teeth in a deformation-based reconstruction process. Based on the resulting 3D model of the dentition, the information relevant to the medical staff can then be calculated.

© Fraunhofer IGD
The tooth contours and numbering from the five routinely taken photographs are used to automatically reconstruct a 3D model of the dentition.