Dental Arch Analysis Using Deep Learning for X-Ray-Free Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing non-visible parts of the mouth, such as embedded teeth and jaw bones, are limited by health risks from X-ray exposure and lack precision in predicting future dental changes, leading to delayed diagnoses and unnecessary treatments.
Innovation Solution
A method using deep learning devices, preferably neural networks, to analyze dental arch representations, incorporating a learning base of historical dental structures to determine attributes of both visible and non-visible dental objects, enabling precise prediction of future changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional X-ray or CBCT tomographic acquisition is used to acquire information on non-visible parts of the mouth, then information on embedded teeth, roots and jaw bones can be obtained, but health risks increase due to repeated X-ray exposure
Solution Approach 1:
The patent uses visible teeth and soft tissues as intermediary objects that can be safely scanned with optical scanners. These visible structures serve as proxies to infer information about non-visible structures through machine learning models, avoiding direct X-ray exposure while still obtaining diagnostic information about roots, embedded teeth, and jaw bones
Solution Approach 2:
The patent replaces the X-ray imaging mechanism with an optical scanning mechanism combined with computational analysis. Instead of using ionizing radiation to visualize internal structures, the system uses reflected light to capture surface geometry and employs deep learning algorithms to predict subsurface anatomical features from these optical scans
2Object-affected harmful factors
If X-ray acquisition is performed only when patients complain of pain or observe abnormalities, then health risks are reduced, but diagnosis becomes late and treatment becomes long and complex
Solution Approach 1:
The patent enables preliminary assessment of dental health by using optical scanners to capture and analyze tooth morphology, position, and surrounding soft tissues before any symptoms appear. The machine learning models predict potential issues with non-visible structures based on visible characteristics, allowing early intervention before pain or abnormalities manifest
Solution Approach 2:
The system establishes a feedback loop where optical scan data is continuously analyzed by machine learning models to detect subtle changes in tooth position, morphology, or surrounding tissues that may indicate developing problems. This ongoing monitoring provides early warning signals that trigger further investigation before clinical symptoms appear
3Object-affected harmful factors
If only visible parts of the mouth are examined, then health risks are minimized, but precision in predicting future dental changes is insufficient
Solution Approach 1:
The patent creates a virtual copy or digital model of the patient's dental anatomy from optical scan data. This digital replica includes not only visible structures but also predicted representations of non-visible structures generated by machine learning models. The virtual model can be manipulated and analyzed to predict future changes without exposing the patient to additional radiation
Solution Approach 2:
The system analyzes changes in multiple parameters of visible structures (tooth position, orientation, morphology, spacing) and uses these parameter changes to infer corresponding changes in non-visible structures. By tracking temporal evolution of visible parameters, the system predicts future dental changes with high precision while avoiding repeated X-ray exposure
Data Source
AI summary
Method of analysis of a diagnostic dental representation showing a dental arch of a current patient in several dimensions. The method includes creation of a learning base including more than 1,000 historical dental structures. Each historical dental structure includes a historical dental representation showing an arch of a historical patient in several dimensions and a historical specification containing a value for at least a first attribute relating to a dental object associated with the historical dental representation. The method includes training of at least one deep learning device by use of the learning base. The method includes submission of the diagnostic dental representation to the deep learning device in such a manner that it determines, for the diagnostic dental representation, at least one value for the first attribute.