3D Facial Shape Prediction Using Vector Average Differences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting facial and breast shape changes after orthodontic and surgical treatments lack precision due to subjective correlations between hard and soft tissue movements, leading to inconsistent treatment plans.
Innovation Solution
A method involving arithmetic calculation processing that extracts multi-dimensional feature vectors from pre-treatment data, selects similar case vectors, calculates normalized shape models, and applies vector averages to predict post-treatment shapes, enhancing precision and objectivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If prediction is based on subjective correlation between hard tissue and soft tissue movement, then treatment planning can be performed, but prediction precision deteriorates
Solution Approach 1:
The patent replaces the mechanical/subjective correlation method with an information-processing approach using feature vectors and machine learning algorithms. The system extracts quantitative features from cephalometric images, creates feature vectors representing patient characteristics, and uses these vectors to predict soft tissue changes based on hard tissue movements, eliminating subjective judgment while maintaining high precision
Solution Approach 2:
The patent transforms the prediction approach by changing from using direct spatial correlation parameters to using extracted feature parameters. By extracting multiple features (landmark coordinates, angles, distances) and converting them into feature vectors, the system can capture complex relationships between hard and soft tissues through quantitative parameters rather than subjective visual correlation
2Reliability
If correlation constant is set based on subjective view, then prediction can be performed, but reliability deteriorates
Solution Approach 1:
The system performs self-calibration by automatically learning the relationship between hard tissue movement and soft tissue changes from the extracted feature vectors. Instead of requiring manual setting of correlation constants by clinicians, the system autonomously determines the predictive model parameters through analysis of the feature data, improving both reliability and ease of operation
Solution Approach 2:
The patent incorporates feedback mechanisms where the predicted results are compared with actual post-treatment outcomes, and the model parameters are adjusted accordingly. This feedback loop allows the system to continuously improve its prediction accuracy by learning from actual treatment results, thereby enhancing reliability without complicating the operation
3Measurement precision
If only two-dimensional cephalo image is used, then processing is simple, but measurement precision deteriorates
Solution Approach 1:
The patent transitions from two-dimensional cephalometric images to three-dimensional facial shape data by extracting feature vectors that capture spatial relationships in multiple dimensions. The system processes landmark coordinates, angles, and distances that represent three-dimensional structures, enabling precise measurement of facial shape changes while managing complexity through structured feature extraction
Data Source
AI summary
To have convenient and highly precise prediction of the shape of a human body after a treatment by calculation processing that includes extracting a feature vector Fnew from face data of a patient as an evaluation subject, selecting a plurality of case patients having feature vectors Fpre(i), extracted from the face data of a plurality of previous patients, obtaining pre-orthodontic facial shape models Hpre(i) and a post-orthodontic facial shape models Hpost(i) in which the faces of the selected previous case patients before and after a treatment have been normalized, obtaining a facial shape model Hnew, and obtaining a three-dimensional predicted facial shape model Hprd, by modifying the facial shape model Hnew of the patient as an evaluation subject, using a vector average difference AVEpost−AVEpre between the pre-treatment and post-treatment facial shape models.