During the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), the Tooploox (a Solvd Inc. company) research team introduced a new method for applying Artificial Intelligence to the analysis of three-dimensional dental scans. The objective is to deliver accurate mapping of tooth structure, positioning, and root systems, while minimizing image noise.
The global market for Artificial Intelligence applications in dentistry is projected to grow from $421 million in 2024 to $3.1 billion by 2034. Growth is driven by the adoption of algorithms that support diagnostics, imaging analysis, preventive care, and treatment planning.
Challenges in applying AI in dentistry
Dental diagnostics typically begin with identifying the structures within the jaw. Beyond visual examination, dentists must document each tooth and record the presence of diseases, abnormalities, or patient-specific conditions, creating a perfect use case for AI application. But in this context, it is not that easy.
Because every patient is unique, distinguishing patterns in dental scans for the sake of AI-support in diagnostics is complex. Several factors contribute to this challenge:
- Dense anatomical environment: The jaw contains interconnected structures such as teeth, gums, bone, and nerves.
- Influence of prior conditions: Past extractions, orthodontic treatments, tooth loss from injury, or poor health alter jaw structure and complicate image analysis.
- Natural variation: Teeth differ significantly across individuals, and accepted dental norms are broad. Roots may take irregular forms even in otherwise healthy teeth, which complicates procedures such as extraction when full mapping of shape and position is unavailable.
The application of Artificial Intelligence in dentistry offers measurable advantages, including lowering costs and supporting improved patient outcomes. Effective care depends on accurate knowledge of the patient’s condition. AI can enhance both diagnostics and procedure planning.
These systems do not replace clinical decision-making but provide data that enables practitioners to make more informed choices.
Our research
To advance these benefits, the Tooploox research team, in collaboration with external partners, developed GEPAR3D (GEometric Prior-Assisted Learning for 3D). The method integrates instance detection – such as identifying individual teeth or other jaw components with segmentation, enabling precise classification of elements within three-dimensional scans.
The model processes a three-dimensional scan of the mouth by referencing the statistical position of teeth. It establishes a central point to initiate detection, then maps the presence or absence of each tooth. Once the structure is identified, the system classifies individual elements, such as implants, fractures, or deformities.
Position-based detection preserves anatomical context and reduces the risk of errors caused by irregular features or imaging flaws. It also improves efficiency by narrowing the search to the most probable locations of specific teeth, thereby reducing computational effort.
More details about the algorithm, the approach, and its internal workings can be found in the research paper published on Arxiv.
The effect
The system generates a three-dimensional model of the teeth with structure and shape clearly defined. By filtering out noise and irrelevant background data, it provides clinicians with a precise view of dental anatomy, including root structures, their interconnections, and positioning relative to other teeth.
GEPAR3D outperforms leading methods in tooth localization and segmentation. Its potential applications include enhancing orthodontic treatment planning and improving the assessment of root resorption.
The next step may be training the algorithm to work with children’s teeth – as the current version works with adult teeth only. Also, the healthcare professional still needs to oversee the outcomes.
The research was delivered by a team consisting of Tomasz Szczepański, Szymon Płotka, Michal K. Grzeszczyk, Arleta Adamowicz, Piotr Fudalej, Przemysław Korzeniowski, Tomasz Trzciński, and Arkadiusz Sitek, representing Solvd; the Sano Centre for Computational Medicine, Cracow, Poland; Jagiellonian University, Cracow, Poland; Jagiellonian University Medical College, Cracow, Poland; the Warsaw University of Technology, Warsaw, Poland; Research Institute IDEAS, Warsaw, Poland; and Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA.
The research will be presented at the upcoming 28th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2025), which will be held from Tuesday, September 23rd, to Saturday, September 27th, 2025, in Daejeon, Republic of Korea. More details about the conference can be found on the event website.
