Lesion Image Subtraction Using AI-Synthesized Contrast
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lesion detection methods, particularly in medical imaging, face challenges such as poor signal-to-noise ratio, difficulty in distinguishing image contrast, and reduced accuracy due to human fatigue, especially for small lesions and those not visible in all imaging modalities, and the limitations of contrast agents.
Innovation Solution
A method combining deep learning and silhouette subtraction techniques using two imaging modalities with different sensitivities to enhance lesion visibility by synthesizing images and performing a subtraction process, enhancing contrast-to-noise ratio without the need for contrast agents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning is used to detect lesions automatically when image quality is satisfied, then detection accuracy is improved, but detection accuracy decreases when image contrast is degraded
Solution Approach 1:
The patent applies preliminary action by synthesizing a contrast-enhanced image using deep learning before the actual lesion detection process. The synthesis module generates a predicted contrast-enhanced image from the original image, preparing enhanced contrast information in advance. This allows the lesion detection to benefit from improved contrast without actually degrading the original image quality, thus maintaining high detection accuracy even when original image contrast is poor.
2Difficulty of detecting and measuring
If contrast agent is used to enhance lesion visibility, then lesion detection capability is improved, but it is limited to patients who are not allergic to the chemicals
Solution Approach 1:
The patent replaces the mechanical/chemical system of contrast agents with an intelligent computational system. Instead of injecting physical contrast chemicals that may cause allergic reactions, the system uses a deep learning synthesis module to computationally generate contrast-enhanced images. This substitution eliminates the safety concerns associated with contrast agents while maintaining their lesion-enhancing effect, making the technique applicable to all patients regardless of allergy status.
3Loss of information
If human radiologists perform intensive image readings, then comprehensive analysis is achieved, but small lesions may be missed due to human fatigue
Solution Approach 1:
The patent introduces an intermediary system between the original medical images and the final diagnosis. The synthesis module acts as an intermediary that generates contrast-enhanced versions of the images, making subtle lesions more visible. This intermediary processing step enhances the information available to radiologists without requiring them to perform more intensive reading, thereby reducing fatigue-related errors while maintaining comprehensive analysis.
Data Source
AI summary
The present invention relates to lesion detection. In order to improve lesion detection, it is proposed to combine the techniques of deep learning and the strategy of silhouette, i.e. subtraction between contrast-enhanced and non-contrast-enhanced, in order to reflect only the difference between the two images related to the lesion.