Supervised Learned Filter for Medical Image Abnormal Component Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image analysis techniques face challenges in accurately generating normal structure images and separating abnormal components, particularly due to variations in abnormal tissue shapes and the reliance on shape-dependent filters, which limits recognition accuracy.
Innovation Solution
A supervised learned filter is used to generate a normal image from input medical images by learning from a set of training images with corresponding supervisor images, allowing for more accurate representation of normal structures and subsequent separation of abnormal components without relying on shape.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If shape-dependent filters are used to separate abnormal components, then the separation process is simplified, but the recognition accuracy deteriorates due to variations in abnormal tissue shapes
Solution Approach 1:
The patent replaces shape-dependent mechanical filtering methods with a learning-based system that automatically adapts to various abnormal tissue shapes. The supervised learned filter substitutes the manual shape-matching approach with an intelligent system that learns from training data, thereby maintaining simplicity while improving recognition accuracy across diverse abnormal shapes.
Solution Approach 2:
The patent changes the filter parameters dynamically based on learned characteristics from training images. Instead of using fixed shape-dependent parameters, the system adjusts filter parameters according to the specific characteristics of each abnormal component, enabling accurate separation regardless of shape variations while keeping the process automated and simple.
2Adaptability or versatility
If AAM (Active Appearance Models) are used to generate normal structure images, then the normal structure can be represented, but the accuracy of the generated normal structure image deteriorates
Solution Approach 1:
The patent introduces a feedback mechanism where the system learns from the differences between input images and generated normal structure images. By using supervisor images as ground truth and continuously refining the filter through supervised learning, the system improves the accuracy of generated normal structure images while maintaining the adaptability to represent various normal anatomical structures.
Solution Approach 2:
The patent performs preliminary learning using training images and supervisor images before actual abnormal component separation. This preliminary action of supervised learning pre-configures the filter with accurate normal structure characteristics, ensuring high accuracy when generating normal structure images during actual operation while maintaining versatility across different subjects.
Data Source
AI summary
A normal image representing a normal structure of a predetermined structure in an input medical image is generated with higher accuracy. Further, an abnormal component in the input medical image is separated with higher accuracy. A supervised learned filtering unit inputs an input image representing a predetermined structure to a supervised learned filter to generate an image representing a normal structure of the predetermined structure. The supervised learned filter is obtained through a learning process using supervisor images, each representing a normal structure of the predetermined structure in a subject (individual), and corresponding training images, each containing an abnormal component in the corresponding subject (individual). Further, a difference processing unit separates an abnormal component in the input image by calculating a difference between the input image and the image representing the normal structure.