Medical Image Training Data Creation via Lesion Region Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image processing systems struggle to effectively create training data for computer-aided detection and diagnosis systems, particularly in accurately aligning and associating lesion region information with medical images.
Innovation Solution
A medical image data creation apparatus and method that acquires two medical images of the same spot, aligns them, creates lesion region information, and associates this information with the first image to generate medical image data for training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple medical images of the same spot are acquired and aligned, then the accuracy of lesion detection is improved, but the complexity of data processing increases
Solution Approach 1:
The patent segments the complex task of training data creation into distinct modules: image acquisition, alignment processing, lesion region detection, and data association. Each module handles a specific function, making the overall complex process manageable and systematic. The alignment unit separately processes positional relationships, while the lesion region information creation unit focuses on identifying pathological areas.
Solution Approach 2:
The patent introduces an alignment unit as an intermediary component that bridges the gap between multiple acquired images and the lesion detection process. This alignment unit processes the positional relationships between images before they are used for training, serving as a mediator that simplifies the integration of multiple image sources while maintaining detection accuracy.
2Reliability
If lesion region information is created from aligned images, then the quality of training data is improved, but the time required for data creation increases
Solution Approach 1:
The patent performs alignment processing as a preliminary action before creating lesion region information. By pre-aligning the multiple medical images and establishing their positional relationships in advance, the system prepares the data structure needed for accurate lesion detection. This preliminary alignment reduces the computational burden during the actual lesion detection phase.
Solution Approach 2:
The system automatically creates lesion region information by processing the aligned images through the lesion region information creation unit, eliminating the need for manual annotation. The apparatus serves itself by autonomously identifying lesion regions and associating them with the corresponding image data, significantly reducing the time required compared to manual methods.
3Productivity
If multiple images are aligned and associated with lesion information, then the effectiveness of CAD system training is improved, but the computational resources required increase
Solution Approach 1:
The patent extracts and processes only the essential information from multiple medical images - specifically the alignment data and lesion region information. Rather than processing all image data in full detail, the system extracts the critical components needed for CAD training, reducing unnecessary computational overhead while maintaining training effectiveness.
Solution Approach 2:
The patent transforms the problem from processing raw image pixels to processing structured data relationships. By creating an associative structure that links lesion region information with aligned image data, the system moves from dimensional image processing to a more efficient data relationship processing model, reducing computational resource requirements.
Data Source
AI summary
A medical image data creation apparatus for training acquires a first medical image, acquires a second medical image that is different from the first medical image and is obtained by photographing a substantially same spot as in the first medical image, creates lesion region information concerning a lesion part in the second medical image, and creates medical image data for training in which the first medical image and the lesion region information are associated with each other.