3D Modeling Using Simple Intraoral Cameras: A Research Study
To evaluate the feasibility of 3D modeling of 2 types of typodonts from video files using stereo photogrammetry technique assisted by AI technology.
2 types of typodont's mandibles were used, one with full teeth and the other one with 4 teeth removed. Video files were created using iPhone 15, commercially available intra oral cameras (IOC) Qawachh and Waldent 5 Megapixel Intraoral cameras. 5 scans were obtained from both arches using the 3 different cameras mentioned. Using the video files 3D models were generated by stereophotogrammetry technology using AI assistance. These 3D models were compared with the gold standard Desktop scanner produced 3D models.
Reasonable high quality 3D meshes were generated by iPhone video files (92% overlap) followed by Waldent 5 Megapixel (90% overlap) and a lesser quality by Qawachh (85% overlap) intra oral cameras.
Video files from intraoral cameras can be successfully converted into 3D models with reasonable accuracy using stereophotogrammetry with AI assistance.
If we can improve the accuracy even more in generating 3D models of intra oral structures using simple intraoral cameras, we will be able to use them in underserved areas because of the very low cost of this technology.
The advent of CAD/CAM digital technology has transformed dental practices, particularly in the area of 3D modeling for diagnosis, treatment planning, and prosthesis fabrication. Traditional methods of creating physical impressions, though reliable, are increasingly being supplemented or replaced by digital solutions. Among these the latest intraoral scanners and desktop scanners have set the standard for generating accurate 3D representations of dental structures. However, these devices often come with significant cost, learning curve and training requirements, limiting their accessibility in many settings.
To address these challenges, this study explores the use of cost-effective alternatives, AI assisted stereophotogrammetry using simple dental cameras, for 3D dental modeling. Stereophotogrammetry relies on capturing overlapping 2D images to reconstruct detailed 3D models, while simple dental cameras provide a more affordable option for image acquisition. By leveraging the latest technology, the research aims to assess the feasibility of economically accessible solutions for dental professionals.
Ali Saghiri et al showed that 2D images from still pictures can be used to produce accurate 3D models of the mandible. However, they used ReCap photo a 3rd party software, which is no longer a freeware and also it is difficult to integrate it into routine workflow. Using our own AI algorithms, we were able to bypass this process.
The projected light can be red, white or blue light with different patterns such as a line, point or a projected mesh.
In passive light technology such as stereophotogrammetry, the estimate of all three coordinates is calculated through computer algorithms using software only. Algorithms are used in estimating the camera positions and generating a point cloud for 3D model creation.
| Feature | Hardware-Heavy IOS | Software-Dependent Technology |
|---|---|---|
| Cost | Up to $40,000 + maintenance fees | Significantly lower (uses existing cameras) |
| Size/Weight | Bulky (up to 1lb) | Lightweight, smaller form factor |
| Training Required | Extensive training, steep learning curve | Minimal training, intuitive operation |
| Ecosystem | Limited to specific vendor | More flexible, device-agnostic |
The Ultrassist Typodont Teeth Model was chosen for this study due to its accurate representation of human dental anatomy and versatility in replicating clinical scenarios.
All scans were performed under controlled ambient lighting at 1000 lux to ensure consistency.
Systematic path beginning with occlusal surface, followed by buccal and lingual surfaces in a sequential manner.
Each typodont model was scanned five times with each device to evaluate consistency and accuracy.
The reconstruction process involved generating both sparse and dense point clouds to represent the 3D geometry of the scanned object.
D = √∑(f₁ᵢ-f₂ᵢ)²
Where f₁ᵢ and f₂ᵢ are feature descriptors from two images.
E = ∑‖xᵢⱼ - PᵢXⱼ‖²
Where xᵢⱼ are observed 2D points, Pᵢ is the projection matrix for camera i, and Xⱼ are 3D points.
T(u,v) = I(x,y)
This ensured accurate surface details and high-quality visual output.
GPU-accelerated algorithms enhanced feature detection, point cloud generation, and mesh reconstruction, enabling efficient processing of large datasets.
iPhone 15
Overlap with reference model
Waldent 5MP
Overlap with reference model
QAWACHH
Overlap with reference model
AI has significantly helped in various steps of our process:
This study demonstrates that video files captured with simple intraoral cameras can be effectively converted into accurate 3D models using stereophotogrammetry techniques enhanced by AI algorithms.
We believe that we can improve the accuracy to 99%+ by:
The potential impact of this technology is significant, particularly for underserved areas where access to expensive dental equipment is limited. By providing a low-cost alternative for 3D modeling, this approach could democratize access to advanced dental care technologies.
Learn more about our ongoing research in AI-powered dental imaging solutions.