Neural Network Dental Movement Tracking From Mobile Photos
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for remotely monitoring dental situations during orthodontic treatment require high computing resources, 3D scanning, and patient visits, which are costly and inconvenient.
Innovation Solution
A method involving training a neural network with a historical learning base of dental organ images and spatial attributes to analyze dental situations using simple photos from a mobile phone, eliminating the need for 3D scanning and patient visits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D scanning and initial model creation are used to ensure accurate dental monitoring, then measurement precision is improved, but device complexity and loss of time increase
Solution Approach 1:
The patent uses neural networks to create a virtual copy of the dental situation from simple photos, replacing the need for physical 3D scanning. The neural network learns from historical data to generate accurate spatial information from 2D images, achieving the same measurement precision without the complex hardware
Solution Approach 2:
The patent replaces the mechanical 3D scanning system with a neural network-based image analysis system. Instead of using physical scanners to capture dental geometry, the system uses trained neural networks to extract spatial information from photographs, substituting mechanical measurement with computational analysis
2Measurement precision
If 3D scanning and initial model creation are used to ensure accurate dental monitoring, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary training of neural networks on historical data before actual use. Once trained, the neural networks can quickly analyze new patient photos without requiring time-consuming 3D scanning procedures. The heavy computational work is done in advance during training, making the actual analysis fast
3Productivity
If neural network training with historical learning base is used to analyze dental situations, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent trains neural networks in advance using historical learning bases containing numerous examples of dental situations. This preliminary training allows the models to quickly analyze new cases without requiring complex real-time computations. The computational heavy lifting is performed during training, not during actual dental analysis
Solution Approach 2:
The patent creates a virtual library of pre-processed dental data from historical cases. The neural networks learn patterns from these copied historical examples, enabling them to quickly recognize and analyze similar situations in new patients without repeating complex analysis procedures
4Reliability
If traditional monitoring methods requiring patient visits are used, then reliability is improved, but ease of operation deteriorates
Solution Approach 1:
The patent enables patients to perform self-monitoring by taking their own photos and having the neural networks automatically analyze their dental situations. The system serves itself by automatically comparing current photos with historical data to detect changes, eliminating the need for professional visits while maintaining monitoring reliability
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Disclosed is a method for training a neural network intended to analyse a dental situation of an updated patient, the method comprising the following steps: A) creating a historical training database relating to a dental body and to a spatial attribute associated with the dental body, the historical training database comprising more than 1000 historical records, each historical record, relating to a respective historical patient, comprising: - a set of historical images all representing the dental body in the historical patient, referred to as the "historical dental body"; and - an item of spatial information comprising, for the historical patient, a set of values for the spatial attribute, referred to as "historical spatial information"; B) training the neural network, by providing it with the sets of historical images as input and the historical spatial information as output.