Dental Object Detection Using Multi-View Projection and Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting teeth in oral scan images and registering them with CT images are prone to errors due to ambiguous boundaries between teeth and gums, leading to inaccurate and time-consuming processes.
Innovation Solution
A method involving the extraction of regions of interest from oral scan images, generation of learning models trained on multiple directions of tooth objects, and detection of objects in these images to derive accurate reference points for image registration, improving detection efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If curvature information is used to detect teeth in oral scan images, then the detection process can be performed, but detection errors occur frequently due to ambiguous boundaries between teeth and gums and similar curvature information
Solution Approach 1:
The patent segments the tooth detection problem into multiple viewpoints by generating 2D projection images from different angles (mesial, distal, buccal, lingual, occlusal, and apical views). This segmentation allows the system to analyze tooth characteristics from multiple perspectives, making it easier to distinguish tooth boundaries from gum tissue and improving detection accuracy and reliability.
2Measurement precision
If conventional methods are used to perform image registration between oral scan images and CT images, then image registration can be performed, but the speed is reduced due to large computational load
Solution Approach 1:
The patent extracts key reference points (tooth apex points and centroid points) from the 2D projection images and uses these extracted points for image registration. By taking out only the essential registration features rather than processing entire images or comparing all vertices, the system achieves accurate registration while significantly reducing computational load and improving processing speed.
3Ease of operation
If manual markers or vertex comparison is used for image registration, then image registration can be performed, but accuracy is reduced due to inaccurate characteristics of manual markers and vertices
Solution Approach 1:
The patent enables the system to automatically identify and extract tooth reference points (apex and centroid) from the 2D projection images without requiring manual marker placement. The learning model automatically detects and marks these critical points, making the system self-sufficient for the registration process while achieving high accuracy through consistent, objective point selection based on tooth geometry.
4Measurement precision
If all vertices in each image are compared for image registration, then comprehensive comparison can be performed, but the computational load increases significantly
Solution Approach 1:
The patent applies local quality by focusing computational resources on specific critical regions - the apex points and centroid points of teeth - rather than uniformly processing all vertices. By identifying and analyzing only these locally significant points that best represent tooth position and orientation, the system achieves accurate registration with minimal computational energy expenditure.
Data Source
AI summary
An image registration method using a dental object, comprises a generation step of generating a plurality of reference points spaced apart from each other in an oral scan image of a registration target and a computed tomography (CT) image of a registration target, respectively; and a step of registering the oral scan image of the registration target and the CT image of the registration target by using the reference point of the oral scan image (a first reference point) and the reference point of the CT image (a second reference point), wherein the first and second reference points comprise a reference point for one frontmost tooth in the front teeth area and a reference point for two teeth on both rearmost sides of the back teeth area, and wherein the first reference point is derived from an object which is a simplified shape of the teeth.