3D Medical Image Anonymization via Background Randomization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing accessibility and sharing of medical X-ray and MR images raise concerns about patient privacy, as direct volume rendering techniques can inadvertently or maliciously reveal recognizable body surfaces, potentially leading to abuse and discouraging patients from consenting to medical imaging or clinical studies.
Innovation Solution
An image processing method that segments 3D image data into object and background regions, applying randomization to the background region to transform the image data set, making the body surface unrecognizable while retaining full resolution for internal organs, with the option to reverse the randomization for medical diagnosis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If direct volume rendering is applied to 3D medical image data, then the ability to visualize internal structures is improved, but the risk of revealing recognizable body surfaces increases
Solution Approach 1:
The patent segments the 3D image data into different regions: an inner core region containing internal structures and an outer layer region representing the body surface. This segmentation allows selective processing where the inner core retains full resolution for diagnostic purposes while the outer layer is randomized to protect patient privacy.
Solution Approach 2:
The patent applies different quality characteristics to different regions of the 3D data. The inner core region maintains high resolution and original image quality for medical diagnosis, while the outer layer region applies randomization to eliminate recognizable surface features. This local differentiation resolves the contradiction by preserving internal information while protecting external privacy.
2Object-affected harmful factors
If randomization is applied to the entire 3D image data set, then patient privacy is protected, but the utility for medical diagnosis is reduced
Solution Approach 1:
The patent divides the 3D image data into an inner core region and an outer layer region, applying randomization only to the outer layer while preserving the inner core. This segmented approach ensures privacy protection at the surface level while maintaining diagnostic quality within the body interior.
Solution Approach 2:
The patent implements local quality differentiation by applying randomization operations selectively to the outer layer region only, while leaving the inner core region unchanged. This ensures that privacy protection is applied where needed (at the surface) while diagnostic information remains intact in the interior.
3Object-affected harmful factors
If the outer layer region is randomized, then body surface recognizability is eliminated, but the complexity of image processing increases
Solution Approach 1:
The patent segments the processing task into two distinct operations: segmentation of the 3D data into inner core and outer layer regions, and randomization of only the outer layer. This segmentation simplifies the overall process by limiting randomization to a specific region rather than processing the entire data set.
Solution Approach 2:
By applying randomization locally only to the outer layer region rather than the entire 3D data set, the patent reduces the computational complexity and processing time required. The inner core region bypasses the randomization operation, maintaining its original image quality and processing efficiency.
Data Source
AI summary
Image processing method or apparatus (IP) to transform a 3D image data set (DS) into a visually protected one (DSX). The 3D image set includes an object region (OR) and a background region (BR) that defines s silhouette of an imaged object (P). An inadvertent or malicious direct volume rendering of the silhouette (IF) of the object is prevented by applying a randomization operation to at least the background region (BR).