Cross-Modality Feature Mapping for CADx Case Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current case-based computer-aided diagnosis (CADx) systems are limited in their ability to retrieve similar cases across different imaging modalities, hindering clinicians' ability to make accurate diagnoses by not allowing for effective comparison of tumors or lesions from various imaging techniques like CT, MRI, and ultrasound.
Innovation Solution
A cross-modality CADx system that maps feature relationships between images from one modality to another, allowing for the retrieval and simultaneous display of similar cases across multiple imaging modalities by using techniques like Factor Analysis and polynomial functions to establish feature ratios and mappings, enabling comparison and diagnosis based on a single modality image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If case-based CADx systems retrieve similar cases only from the same imaging modality, then the system simplicity is maintained, but the diagnostic accuracy and clinical utility are limited due to inability to compare across different imaging techniques
Solution Approach 1:
The patent introduces a feature mapping module as an intermediary that translates features from one imaging modality to another modality's feature space. This mediator enables cross-modality case retrieval by converting CT features to MRI feature representations (and vice versa) using pre-established mapping relationships, thus allowing clinicians to compare cases across different imaging techniques without directly integrating complex multi-modality systems
Solution Approach 2:
The system implements a universal feature space that can represent cases from multiple imaging modalities (CT, MRI, ultrasound, etc.) through a common feature representation framework. The feature mapping mechanism enables the system to handle multiple modalities uniformly, allowing a single query interface to retrieve relevant cases across all supported modalities, thereby achieving multi-functionality without proportionally increasing system complexity
2Adaptability or versatility
If cross-modality feature mapping is implemented, then the ability to retrieve and compare cases across different imaging modalities is enabled, but the computational complexity and data processing requirements increase
Solution Approach 1:
The system performs feature mapping relationships in advance by pre-processing and establishing mapping tables between different imaging modalities during system initialization or offline processing. This preliminary action stores pre-computed feature correspondences (e.g., CT density mapping to MRI signal intensity) that can be quickly applied during online case retrieval without real-time computational overhead, thus reducing online processing complexity
Solution Approach 2:
The patent transforms the problem of cross-modality comparison by changing the parameter representation space. Instead of comparing raw imaging parameters from different modalities directly, the system maps all modality-specific parameters to a unified feature space with standardized representations, enabling consistent comparison across modalities through parameter transformation rather than complex multi-parameter integration
Data Source
AI summary
A system and method for cross-modality case-based computer-aided diagnosis comprises storing a plurality of cases, each case including at least one image of one of a plurality of modalities and non-image information, mapping a feature relationship between a feature from images of a first modality to a feature from images of a second modality, and storing the relationship.