Medical Image Training With Geometric Shape Composite Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning systems for medical images, particularly ultrasound images, struggle to accurately recognize images with superimposed geometric shapes due to insufficient training data, and preparing such data is time-consuming and difficult.
Innovation Solution
A machine learning system that includes an image database, a geometric-shape database, and a processor to select and combine medical images with geometric shapes, generating composite images for efficient training of a learning model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of combined medical images are prepared for machine learning training, then the recognition accuracy of medical images with geometric shapes is improved, but the preparation time and difficulty increase significantly
Solution Approach 1:
The system performs preliminary actions by automatically generating composite images with geometric shapes superimposed on medical images before training begins. The image generation unit creates multiple variations of geometric shapes (rectangles, circles, triangles, polygons) at different positions, sizes, and orientations on training images, so that when training starts, the model is already exposed to diverse geometric shape configurations without manual intervention.
Solution Approach 2:
The system creates copies of medical images with various geometric shapes superimposed on them. The image generation unit generates multiple copies of the same medical image, each with different geometric shapes placed at different locations and orientations. This allows the training model to learn from multiple variations without requiring multiple original medical images, significantly reducing data preparation time while maintaining training effectiveness.
2Adaptability or versatility
If geometric shapes are superimposed on medical images for training, then the model can recognize images with geometric shapes, but the model may mistakenly learn geometric shapes as features and fail to recognize images without geometric shapes
Solution Approach 1:
The system applies local quality by placing geometric shapes at specific locations (regions of interest) on medical images rather than uniformly across the entire image. The image generation unit superimposes geometric shapes on specific organs or anatomical structures identified in the medical images, allowing the model to learn that geometric shapes are annotations indicating important regions rather than inherent features of the entire image. This localized approach helps the model distinguish between annotation markers and actual anatomical features.
Solution Approach 2:
The system uses partial action by generating composite images with geometric shapes only for certain training samples or for specific regions within images. Rather than adding geometric shapes to every possible location in every training image, the system strategically places them in meaningful regions, allowing the model to learn both with and without geometric shape contexts, thereby maintaining reliability for images without geometric shapes while adapting to images that do contain them.
3Quantity of substance
If manual preparation of combined medical images is performed, then the training data can be created, but the process is troublesome and not easy to execute
Solution Approach 1:
The system implements self-service by enabling automatic generation of composite training images without requiring manual intervention. The image generation unit automatically retrieves medical images from the image database, superimposes various geometric shapes at different positions and orientations, and saves the generated composite images back to the database. This automated self-service process eliminates the need for researchers or technicians to manually create thousands of composite images, making the data preparation process trivial and easily executable.
Solution Approach 2:
The system replaces the mechanical manual process of image composition with an automated computational system. Instead of manually placing geometric shapes on medical images using image editing software, the patent employs a computer-based image generation unit that automatically performs the superposition operation through programming. This substitution of mechanical manual labor with automated computational processing dramatically increases efficiency and ease of operation while generating large quantities of training data.
Data Source
AI summary
There are provided a machine learning system, a learning method, and a program that can facilitate learning of a large number of combined medical images. A machine learning system (10) includes: an image database (14) that stores a plurality of medical images; a geometric-shape database (16) that stores geometric shapes to be superimposed on the medical images; a processor (22); and a learning model (30), in which the processor (22) is configured to perform a selection accepting process of accepting selection of a medical image from among the medical images stored in the image database (14) and selection of a geometric shape from among the geometric shapes stored in the geometric-shape database (16), a geometric-shape combining process of combining the selected medical image and the selected geometric shape and generating a composite image, and a training process of making the learning model perform learning by using the composite image.