Machine Learning Tooth Movement Prediction for Aligner Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current aligner treatment methods suffer from imprecision in predicting tooth movements, leading to deviations between planned and actual tooth positions, necessitating frequent adjustments and additional scans, which increases treatment duration and resource consumption.
Innovation Solution
A machine learning model, such as a neural network, is trained to predict deviations between planned and actual tooth movements using planned tooth movements output by existing software and actual tooth movements after alignment, allowing for improved tooth movement planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current aligner software is used to plan tooth movements, then the treatment plan can be created efficiently, but the prediction accuracy of tooth movements is imprecise (only 50% accuracy)
Solution Approach 1:
The patent replaces traditional mechanical/software-based tooth movement prediction models with a machine learning model. The ML model processes complex dental scan data and predicts actual tooth movements with higher accuracy than conventional software, resolving the contradiction between prediction precision and system complexity by introducing intelligent algorithms that handle complexity automatically.
Solution Approach 2:
The patent changes the parameters used for tooth movement prediction from traditional geometric and mechanical parameters to data-driven parameters derived from machine learning models. This involves transforming input data formats and prediction outputs into a new parameter space that captures non-linear relationships in tooth movement, thereby improving accuracy without requiring explicit mechanical models.
2Reliability
If traditional aligner treatment is performed without periodic adjustments, then the treatment process is simple, but teeth often do not move as planned requiring frequent check-up appointments
Solution Approach 1:
The patent applies preliminary action by using machine learning models to predict actual tooth movements before the treatment begins. The system pre-calculates compensation factors and adjusts the aligner treatment plan in advance to account for expected deviations, ensuring teeth move as planned without requiring frequent mid-treatment adjustments or check-up appointments.
Solution Approach 2:
The patent implements feedback mechanisms where actual tooth movements are continuously monitored and compared against predicted movements. The machine learning model processes this feedback data and automatically refines subsequent treatment plans, creating a closed-loop system that adapts to real-world conditions and maintains reliability without extending treatment duration through multiple adjustment cycles.
3Manufacturing precision
If multiple supplementary sets of aligners are fabricated to compensate for deviations, then tooth alignment can be achieved, but treatment complexity and resource consumption increase
Solution Approach 1:
The patent applies preliminary action by using machine learning models to predict actual tooth movements before the treatment begins. The system pre-calculates compensation factors and adjusts the aligner treatment plan in advance to account for expected deviations, ensuring teeth move as planned without requiring frequent mid-treatment adjustments or check-up appointments.
Solution Approach 2:
The patent implements feedback mechanisms where actual tooth movements are continuously monitored and compared against predicted movements. The machine learning model processes this feedback data and automatically refines subsequent treatment plans, creating a closed-loop system that adapts to real-world conditions and maintains reliability without extending treatment duration through multiple adjustment cycles.
Data Source
AI summary
A method of training a machine learning model to determine tooth movements for a process of tooth alignment of a subject. The method comprises receiving, in a computer memory, for a plurality of subjects: planned tooth movements for a tooth alignment process for aligning one or more teeth of the respective subject; and actual tooth movements for said one or more teeth of the respective subject after a tooth alignment process has been performed. The method further comprises training one or more machine learning models at least in part using the planned tooth movements. The one or more machine learning models are arranged to extract one or more predicted deviations based at least in part on the planned tooth movements. The or each predicted deviations represent differences between the planned tooth movements and expected tooth movements if the planned tooth movements were used in a tooth alignment process.