Unified Medical Image and Report Retrieval System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for medical image retrieval lack the capability to combine semantic search in structured data with image similarity search, requiring users to remember similar medical cases by patient name for effective searches.
Innovation Solution
A method and device that analyze images and reports to detect structures and text passages, map them to unique resource identifiers, compute features, and aggregate search results from both semantic and image similarity queries to provide a comprehensive final result list.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If only semantic search or only image similarity search is used, then the search system is simple to implement, but the search capability is weak and users must remember patient names
Solution Approach 1:
The patent combines semantic search (based on structured data and ontologies) with image similarity search (based on pixel intensities and low-level features) into a unified retrieval system. The query processing module integrates both search types to provide comprehensive search results, eliminating the limitation of using only one search method while maintaining manageable system complexity through modular architecture.
Solution Approach 2:
The retrieval system is designed to handle multiple query types simultaneously - semantic queries using structured data, image similarity queries using visual features, and combined queries. The system provides a universal interface that accepts different query formats and routes them to appropriate processing modules, enabling flexible search capabilities without requiring separate systems for each search type.
2Adaptability or versatility
If combined semantic and image similarity queries are implemented, then search flexibility and accuracy improve, but system complexity increases
Solution Approach 1:
The system segments the query processing into distinct modules: a query processing module that handles semantic queries using structured data and ontologies, an image retrieval module that handles similarity queries using visual features, and a result integration module that combines results from both modules. This segmentation allows each module to specialize in its function while the overall system provides combined search capabilities.
Solution Approach 2:
The patent introduces intermediary components including a query processing module that translates user queries into both semantic and image-based search parameters, and a result integration module that mediates between the two search results. These intermediaries simplify the interaction between different search types and provide a unified interface for users, reducing the perceived complexity while enabling versatile query capabilities.
Data Source
AI summary
Methods and device are disclosed for data retrieval. At first images and reports are analyzed by respective parser units to detect both structures and text passages that are related to respective structures and text passages of a knowledge data-base. The detected structures and text passages are stored together with a unique resource located that identifies the respective structure and/or text passage at the knowledge database in a semantic annotation database. In addition a feature extraction can be performed to provide specific features of the images and/or regions of the images, whereby the features are stored in an image feature database. Finally an input query can ask questions that are used to provide a result to the query based on the semantic annotation database and the image feature database. The methods and devices may be used for data preparation and data retrieval of medical images and associated medical reports.