Machine-Learned Texture Features for Early Therapy Response Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing tumor therapy response rely on geometric measures, which are slow to change, leading to prolonged therapy exposure and higher costs, while texture changes offer earlier insights but are time-consuming and difficult to analyze manually.
Innovation Solution
A computer-aided system using machine learning with data-driven texture features, such as Independent Subspace Analysis, to predict therapy response from medical images, allowing for early assessment with potentially fewer scans.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If geometric measures are used to assess therapy response, then measurement reliability is improved, but assessment time increases and therapy duration is prolonged
Solution Approach 1:
The patent changes the measurement parameter from geometric properties (size, shape) to textural properties (intensity variations, patterns) of the lesion. Texture features such as those extracted by Independent Subspace Analysis capture tissue composition changes that occur earlier than geometric changes, enabling earlier reliable assessment of therapy response without waiting for size reduction.
Solution Approach 2:
The patent replaces manual texture analysis with an automated computer-aided system using machine learning classifiers. The system automatically extracts texture features from medical images and predicts therapy response, eliminating the time-consuming manual analysis while maintaining or improving assessment reliability through consistent, objective measurements.
2Productivity
If texture features are used for early therapy response assessment, then assessment speed is improved, but analysis complexity increases
Solution Approach 1:
The patent implements self-service through automated machine learning classifiers that automatically extract texture features and predict therapy response without requiring manual analysis. The system performs feature extraction, classification, and prediction autonomously, transforming complex texture analysis into a streamlined automated process that improves assessment speed while managing complexity through algorithmic automation.
3Measurement precision
If multiple scans are performed to monitor geometric changes, then measurement precision is improved, but cost and radiation exposure increase
Solution Approach 1:
The patent changes from monitoring geometric parameters (size, volume) to textural parameters (intensity patterns, texture features) that provide equivalent or superior measurement precision for therapy response assessment. Texture changes occur earlier and can be detected with fewer scans, reducing radiation exposure and therapy costs while maintaining precision through sensitive texture feature extraction.
4Measurement precision
If manual texture analysis is performed to assess early therapy response, then prediction accuracy is improved, but ease of operation deteriorates
Solution Approach 1:
The patent replaces manual texture analysis with an automated computer-aided system using machine learning classifiers. The system automatically extracts texture features from medical images, applies Independent Subspace Analysis or other feature extraction methods, and predicts therapy response. This substitution maintains high prediction accuracy while dramatically improving ease of operation by eliminating the need for manual analysis expertise and time.
Data Source
Figure 1~2
Figure 3~6
Figure 4~5C
AI summary
For therapy response assessment, texture features are input for machine learning a classifier and for using a machine learnt classifier. Rather than or in addition to using formula-based texture features, data driven texture features are derived from training images. Such data driven texture features are independent analysis features, such as features from independent subspace analysis. The texture features may be used to predict the outcome of therapy based on a few number of or even one scan of the patient