AI dental segmentation: from CBCT to a 3D anatomical model
AI dental segmentation identifies regions in a CBCT volume so they can be represented as separate structures. It is more than adding color to a scan: the process assigns parts of the volume to structures and can turn those selections into surfaces. The resulting 3D model is derived information, not a replacement for the source scan. S01 S02
ASTHECA's current service focuses on demonstration-only 3D representations, with support for professionals in the United States and Mexico. Segmentation is the technology discussed here; the service is not an automated diagnosis or a clinical treatment-planning system.
A 3D view does not necessarily mean individual structures are segmented
A reconstruction can display a volume without making every tooth a separate object. Segmentation identifies regions; a surface can then be generated from that selection. The source volume, the labeled regions, and the displayed mesh should therefore be distinguished. 3D Slicer's documentation describes these different representations of a segmentation. S02
For a dentist reviewing a demonstration, this distinction has a practical consequence: being able to rotate a jaw does not necessarily mean being able to select an individual incisor. Conversely, selecting that incisor does not establish that every boundary has been correctly resolved.
Start by asking which structures were represented, which were excluded, and where the available information ends. Those questions reveal more than the lighting and polish of a render.
What artificial intelligence contributes
AI can propose anatomical labels and boundaries using learned patterns. Published CBCT research includes methods for identifying and segmenting individual teeth; separating a tooth from neighboring teeth and surrounding bone is a central challenge. S14
That progress should not be turned into a blanket claim about every service. A favorable research result does not establish that ASTHECA uses the same model, was evaluated on the same data, or achieves the same performance.
A useful technology demonstration makes its own process understandable: what the representation was intended to show, what the segmentation produced, and which presentation changes were made afterward. An attractive animation can support that explanation, but it cannot replace it.
How to examine a dental model without confusing presentation with evidence
Begin with a narrow question: which part of the scan is being represented? Then identify the objects and their spatial context. An overview provides orientation; an isolated view clarifies what has been assigned to a particular structure.
A third view or accompanying note should address the example's limitations. A region may be outside the requested scope, omitted from the model, or difficult to delineate. These situations do not need to be hidden to demonstrate the technology. They need to be explained.
For a clear presentation, distinguish three states: represented, not represented, and uncertain. The last two must not automatically be interpreted as anatomical absence.
What to define before requesting processing
Describe the demonstration purpose and structures of interest rather than asking only for an STL file. A whole-jaw scene and a presentation with independently selectable objects are different requests.
Also agree on the delivery method and accompanying information. A short description can identify the version, included structures, exclusions, and permitted use. This is a recommendation for defining the request, not a promise that every format or feature is available for every study.
ASTHECA's published workflow begins with coordinating a PROCAOps account. Scope and delivery are agreed before payment. WhatsApp is used for coordination, not for sending patient information or scan files. S13
What a demonstration cannot establish
A well-lit mandible does not demonstrate clinical accuracy. A continuous surface does not establish that the source data resolved every boundary. An anatomical label does not turn a mesh into a radiology report.
ASTHECA's current representations are for demonstration only. They must not be used for diagnosis, treatment planning, or guiding procedures. Clinical evaluation requires professional assessment of the original patient data and other relevant information. S13
Frequently asked questions
Does a 3D viewer mean every tooth has been segmented?
Not necessarily. These are different capabilities. Specify whether you need an overview or a representation of particular structures as separate objects.
Does AI remove the need to review the result?
A responsible demonstration still requires checking what is shown and how it is described. Research on another system does not automatically validate a model delivered by ASTHECA.
Can a U.S. dental professional request the service?
ASTHECA serves the U.S. and Mexican markets. Review the dental service scope and coordinate your request before sharing files through the private channel designated by the team.
Explore the next part of the workflow
Read how to review CBCT segmentation quality to examine boundaries and presentation. For delivery requirements, see what to specify in a dental DICOM-to-STL request.
Need a demonstration-only 3D representation of a dental scan? Explore ASTHECA's dental service. Include business information only in your initial inquiry.
Sources and references
- S01. DICOM PS3.3 — Segmentation IOD — DICOM / NEMA.
- S02. 3D Slicer — Segmentations — 3D Slicer.
- S13. ASTHECA — Your scan, in 3D — ASTHECA, U.S. version consulted 2026-09-06.
- S14. A novel deep learning system for multi-class tooth segmentation and classification on cone beam computed tomography. A validation study — Shaheen et al. / research text published by Relu.