Parameterized Time-Dependent Function for Medical Condition Progression Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting the progression of medical conditions, such as geographic atrophy, are limited by the need for curated data collected at specific time points, making them inflexible and less reliable for predicting future intervals beyond the trained time point, especially in clinical applications where patient data acquisition frequencies vary.
Innovation Solution
A method that uses a parameterized time-dependent function, trained with measurement data from sensors, to predict the progression of a condition over time, allowing for flexible prediction at any time point, including future times, by generating and evaluating parameters indicative of condition changes, and updating the model based on comparison results, using techniques like Fourier or Taylor series to encode complex time-dependent behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a model is trained to predict condition progression at specific time points, then prediction accuracy at those time points is improved, but the model becomes inflexible and unreliable for predicting future intervals beyond the trained time points
Solution Approach 1:
The patent transforms the prediction task from predicting discrete time-point values to predicting parameters of a continuous time-dependent function. The model outputs parameters (e.g., coefficients, amplitude, frequency) that define a function capable of generating predictions at any time point, thereby resolving the contradiction between accuracy at specific points and flexibility for future predictions.
Solution Approach 2:
By having the model predict parameters of a universal time-dependent function rather than specific time-point values, the solution achieves multi-functionality. The same model output can be used to predict condition progression at any future time point, making the prediction system adaptable and versatile while maintaining accuracy.
2Quantity of substance
If prediction models are trained on curated data collected at specific time points, then training data availability is improved, but the models become less reliable for clinical applications where data acquisition frequencies vary
Solution Approach 1:
The invention changes the prediction target from specific time-point measurements to parameters of a continuous function. This allows the model to be trained on available curated data while producing outputs that can reliably predict progression at any time point, including irregular intervals common in clinical practice, thereby resolving the contradiction between data availability and clinical reliability.
3Ease of operation
If a model predicts condition progression at fixed future time points, then prediction simplicity is improved, but the ability to provide continuous progression profiles for treatment planning is reduced
Solution Approach 1:
By predicting parameters of a time-dependent function rather than discrete future values, the model maintains simplicity in the prediction step while enabling continuous progression profiling. The function parameters can be evaluated at any time point to generate continuous profiles needed for treatment planning, resolving the contradiction between operational simplicity and adaptability.
Data Source
AI summary
A method of predicting a progression of a condition comprises obtaining measurement data relating to at least one measurement on a subject up to a particular time point, wherein the data is generated based on an output of a sensor configured to perform the at least one measurement in respect of the subject. At least one parameter of a parameterized time-dependent function, wherein the parameterized time-dependent function is dependent on a continuous time value, is generated using a trained model and based on the measurement data, wherein the parameterized time-dependent function is indicative of a predicted progression of a condition of the subject over time after the particular time point. The parameterized time-dependent function is evaluated using the at least one parameter, for at least one time point after the particular time point.