Unsupervised Edge Restoration in Medical Imaging via Auto-Encoder Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for restoring artificially created cleansed edges in medical images, especially after digital subtraction or material substitution, are tedious and require manual annotation, limiting their flexibility and adaptability across different image types and reconstruction techniques.
Innovation Solution
An apparatus and method using a data-driven model, such as auto-encoders or multivariate regressors, to learn the appearance of image edges from unmodified locations, allowing for unsupervised machine-learning-based restoration without manual annotations, and adapting to varying edge appearances across different image types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation methods are used for edge restoration, then restoration accuracy can be achieved, but the process becomes tedious and time-consuming
Solution Approach 1:
The system performs self-training by automatically learning edge appearance characteristics from unmodified regions of the same image, eliminating the need for manual annotation. The data-driven model trains itself using image data from the input image, making the restoration process autonomous and eliminating tedious manual work while maintaining restoration accuracy
Solution Approach 2:
The system copies edge appearance characteristics from unmodified regions (first image contour) to restored regions (second image contour). By learning the appearance of edges from unmodified locations and applying this learned appearance to restored edges, the system achieves accurate restoration without manual intervention
2Ease of manufacture
If fixed restoration techniques are used, then implementation is simple, but adaptability to different image types and reconstruction techniques is limited
Solution Approach 1:
The system adapts to different image types and reconstruction techniques by dynamically adjusting the data-driven model based on the specific image characteristics. The model learns edge appearance parameters from the input image itself, allowing it to adapt to spectral CT images, virtual mono-energy images, and other modalities without requiring fixed restoration techniques for each case
Solution Approach 2:
The data-driven model serves multiple functions across different image types and reconstruction techniques. A single unified approach using machine learning can handle spectral band images, virtual mono-energy images, and other modalities, replacing the need for multiple specialized restoration techniques
Data Source
Figure 1~2C
Figure 3
Figure 4
AI summary
The present invention relates to edge restoration. In order to improve a restoration of the artificially created cleansed edges, an apparatus is proposed to automatically restore image edges after digital subtraction of digital material substitution to optimally resemble image edges in unmodified locations. The appearance of edges is machine-learned in an unsupervised non-analytical way from unmodified locations, and then, after digital suppression or digital material substitution, applied to the artificially created cleansed edges.