Medical Image Processing Apparatus for Specialized Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing technologies face challenges in performing specialized learning on certain features, such as the degree of spiculation in lung tumor malignancy assessment, due to difficulties in enhancing specific image features and quantifying them for effective machine learning.
Innovation Solution
The method involves calculating and integrating new features by creating image groups based on thresholds set using a machine learning discriminator, allowing for accurate specialized learning on conventional features that were previously difficult to process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning is used to automatically extract features from input images, then prediction accuracy is improved, but the ability to incorporate doctor-important features is lost
Solution Approach 1:
The patent segments the feature extraction process into two independent pathways: one for automatic deep learning feature extraction and another for manual feature design based on doctor knowledge. Both pathways process the same input image separately, allowing each to optimize for its specific purpose without interfering with the other.
Solution Approach 2:
The patent merges the outputs of the automatic feature extraction and manual feature design into a unified feature set that is fed into the prediction model. This combination allows the system to leverage both the automatic capabilities of deep learning and the domain knowledge of doctors, achieving both high accuracy and adaptability.
2Adaptability or versatility
If conventional features are used for image processing, then doctor satisfaction is improved through reflected medical knowledge, but prediction accuracy is reduced compared to deep learning features
Solution Approach 1:
The patent separates the processing of conventional features and deep learning features into distinct modules. The conventional feature module extracts features based on medical knowledge (e.g., texture, shape, size), while the deep learning module automatically extracts features from input images. This segmentation allows both feature types to be optimized independently.
Solution Approach 2:
The patent combines conventional features and deep learning features into a comprehensive feature representation that is input to the prediction model. This merging ensures that both doctor-reflected knowledge and automatically extracted patterns are utilized simultaneously, achieving both high accuracy and medical relevance.
3Adaptability or versatility
If feature enhancement processing is applied to enable specialized learning on conventional features, then learning capability is improved, but the complexity of the processing system increases
Solution Approach 1:
The patent divides the feature processing system into separate modules: a conventional feature extraction module, a deep learning feature extraction module, and a feature integration module. This segmentation allows specialized learning to be performed on conventional features through the first module while the second module handles automatic feature extraction, reducing the complexity burden on any single component.
Solution Approach 2:
The patent designs the feature extraction system to serve multiple functions simultaneously: extracting conventional features, extracting deep learning features, and integrating both types of features into a unified representation. This multi-functionality reduces the need for separate specialized processing systems for each task.
Data Source
AI summary
Provided are a medical image processing apparatus and a medical image processing method capable of implementing specialized learning with higher accuracy in a case where the specialized learning is performed on a plurality of conventional features based on knowledge of a doctor. An image processing apparatus according to the present invention includes: an image group conversion unit that calculates a value of a predetermined feature (first feature) for each image constituting an input first image group, selects an image from the first image group on the basis of the value of the feature, and sets the image as an image of a second image group; and a feature extraction unit that extracts a new feature (second feature) by performing learning on the second image group generated by the image group conversion unit using a feature generation network.