Medical Image Semantic Interpretation via Spatial Relationship Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer-aided diagnosis methods using machine learning and neural networks fail to provide a global semantic interpretation of medical images, only recognizing individual objects without describing their relationships, which limits the accuracy of image analysis.
Innovation Solution
A method that determines categories and positions of objects in an image, calculates their relative positional relationships using a dual-spatial-mask method, and inputs these to a probability knowledge network to obtain a semantic interpretation of the relationships between objects, enhancing the recognition of medical images by describing the entire image context.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional machine learning and neural networks are used for image segmentation, then object regions can be identified, but global semantic interpretation of object relationships is lost
Solution Approach 1:
The patent segments the image analysis process into distinct stages: initial object region segmentation using neural networks, extraction of spatial relationships between segmented regions, and integration of these relationships into a comprehensive semantic interpretation. This segmentation allows preserving relationship information while maintaining processing efficiency
Solution Approach 2:
The patent implements a nested structure where object region segmentation results are embedded within a larger semantic interpretation framework. The segmented object regions serve as nested elements within the comprehensive scene understanding, allowing both local object identification and global relationship interpretation to coexist
2Measurement precision
If only individual object recognition is performed, then processing speed is maintained, but diagnostic accuracy decreases due to lack of context
Solution Approach 1:
The patent performs preliminary extraction of spatial relationships between object regions immediately after segmentation, before final diagnostic interpretation. This preliminary action prepares relationship data in advance, enabling faster comprehensive analysis and reducing overall processing time while maintaining diagnostic accuracy
Data Source
AI summary
A method for recognizing an image, a computer product and a readable storage medium are provided. The method includes: determining categories of a plurality of objects in an image to be detected, and a plurality of object regions where the objects are located; determining positions of the object regions in the image to be detected, and sizes of the object regions; determining a relative positional relationship between the objects according to the positions and the sizes of the object regions; and obtaining a semantic interpretation of the relative positional relationship between the objects according to the relative positional relationship between the objects.