Medical Image Segmentation via Intermediary Object Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image processing methods face challenges in efficiently and accurately segmenting target objects that are difficult to localize and detect, especially in medical volume image data, due to issues like image artifacts and imperfect field of view.
Innovation Solution
An image processing device and method that first segments a well-detectable intermediary object, using its spatial relationship to the target object, to facilitate the detection and segmentation of the target object through adapted shape models and Bayesian frameworks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct target object segmentation is performed on medical volume image data, then segmentation can be attempted immediately, but the difficulty of detecting and measuring increases due to image artifacts and imperfect field of view
Solution Approach 1:
The patent introduces an intermediary object as a mediator between the imaging system and the target object. The intermediary object is easier to detect and segment directly from image data, and its detected position is then used to guide the segmentation of the harder-to-detect target object through spatial relationship constraints, effectively reducing the detection difficulty while maintaining segmentation accuracy
Solution Approach 2:
The patent performs preliminary segmentation of an intermediary object before attempting to segment the target object. By first detecting the intermediary object and establishing its spatial relationship to the target object, the system prepares constraint information in advance that guides the subsequent target object segmentation, making the overall process more efficient and accurate
2Reliability
If model-based segmentation with a priori knowledge is used, then segmentation robustness improves, but the device complexity increases due to shape models and constraints
Solution Approach 1:
The patent uses an intermediary object to simplify the application of model-based segmentation. Instead of applying complex shape models and spatial constraints directly to the target object, the system first segments the simpler intermediary object, then uses its detected position to inform the target object segmentation, reducing the immediate complexity while maintaining robustness
Solution Approach 2:
The patent divides the segmentation task into two separate segmentation processes: first segmenting the intermediary object, then segmenting the target object. This division allows each segmentation to be optimized independently, with the intermediary object segmentation being simpler and the target object segmentation benefiting from constraint information, thereby managing overall system complexity
3Productivity
If intermediary object segmentation is performed first, then the search space for target object detection is constrained improving efficiency, but the device complexity increases due to additional segmentation units
Solution Approach 1:
The patent divides the processing system into specialized units: an intermediary object segmentation unit and a target object segmentation unit. This segmentation of functionality allows each unit to be optimized for its specific task, improving overall efficiency while organizing complexity into manageable, modular components
Solution Approach 2:
The system performs preliminary segmentation of the intermediary object to establish constraint information before target object segmentation. This preliminary action reduces the search space and computational requirements for the subsequent target object segmentation, improving efficiency while the modular architecture manages the added complexity
Data Source
AI summary
The present invention relates to an image processing device and a corresponding image processing method for processing medical image data showing at least two image objects, including a segmentation unit for detection and/or segmentation of image objects in said image data. To allow a more accurate and better segmentation of target objects which are hard to localize and detect, it is proposed that the segmentation unit comprises: a selection unit (61) for selecting a target object for detection and/or segmentation and an intermediary object in said image data, which is easier detectable than said target object and for which position information about the spatial relationship to said target object are known, an intermediary object segmentation unit (62) for segmentation of said intermediary object in said image data, a target object detection unit (63) for detection and/or segmentation of said target object in said image data using said segmented intermediary object and said position information about the spatial relationship of said intermediary object to said target object.