AI Model Augmented Data Generation for Clinical Decision Support
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Clinical decision support (CDS) models face performance deterioration due to data trend changes (drift) not addressed effectively by early retraining methods, leading to inaccurate data estimation and decision boundaries when data sets are small.
Innovation Solution
An information processing apparatus generates augmented data by calculating feature vectors from non-drift and drift data sets, creating candidate data that adheres to the drift trend, thereby evaluating the need for retraining the CDS model and improving inference accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If retraining is performed using a model with data distribution similar to drift data set, then model adaptation to new data trends is improved, but decision boundary accuracy deteriorates due to small number of drift data pieces
Solution Approach 1:
The patent applies preliminary action by generating augmented drift data before retraining the model. Instead of directly retraining with limited drift data, the system first synthesizes additional drift data samples that mimic the characteristics of actual drift data. This preliminary data generation ensures that when retraining occurs, the model has sufficient data to accurately determine decision boundaries, thereby resolving the contradiction between adapting to new trends and maintaining decision accuracy.
Solution Approach 2:
The patent uses copying by creating synthetic drift data samples that replicate the statistical properties and patterns of actual drift data. The augmented data is generated to copy the essential characteristics of drift scenarios, allowing the model to learn from these replicated patterns without requiring large amounts of actual drift data. This copying approach enables accurate decision boundary formation while adapting to new data trends.
2Adaptability or versatility
If counterfactual thinking sample generation method is used, then data versatility is improved, but data accuracy deteriorates by generating unrealistic drift scenarios
Solution Approach 1:
The patent applies parameter changes by systematically modifying data parameters within realistic boundaries to generate augmented drift data. Instead of creating completely counterfactual scenarios, the system adjusts parameters such as data distribution characteristics and feature relationships in controlled ways that reflect plausible drift scenarios. This ensures the generated data maintains accuracy while achieving versatility.
Data Source
AI summary
According to one embodiment, an information processing apparatus for applying to an artificial intelligence (AI) model, includes processing circuitry. The processing circuitry is configured to acquire a first data set having a first data trend and a second data set having a second data trend different from the first data trend, the first data set and the second data set being input to the AI model and discriminated by the AI model. The processing circuitry is configured to calculate a first feature vector based on the first data set and a second feature vector based on the second data set. The processing circuitry is configured to generate augmented data having the second data trend based on the first feature vector and the second feature vector.