Anonymized Medical Image Archive via ROI Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image visualization software faces challenges in preserving patient confidentiality while building a network of annotated medical images from various locations, as large file sizes and identifying metadata pose risks to patient identity, and conventional anonymization methods either remove crucial contextual data or fail to efficiently reduce file size.
Innovation Solution
The method involves generating a compressed archive by degrading the medical image outside the region of interest to low resolution, maintaining high resolution within the region of interest, and anonymizing metadata, while preserving contextual data, to create a network of anonymized medical images that cannot identify patients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire medical image is maintained at high resolution, then image quality is preserved, but file size increases and patient identification risk increases
Solution Approach 1:
The patent segments the medical image into two distinct regions: the region of interest (ROI) maintained at high resolution, and the background region downsampled to low resolution. This segmentation allows selective preservation of critical diagnostic areas while reducing overall file size and eliminating identifying features in non-critical areas.
Solution Approach 2:
The patent applies different quality levels to different parts of the image. The ROI maintains high resolution with full diagnostic quality, while the background region is downsampled to low resolution. This local quality approach ensures that critical areas retain necessary detail while non-critical areas contribute minimally to file size.
2Measurement precision
If the entire medical image is maintained at high resolution, then diagnostic accuracy is preserved, but transmission efficiency decreases
Solution Approach 1:
By segmenting the image into ROI and background regions, the patent enables selective transmission of high-resolution data only where diagnostically necessary. The downsampled background region significantly reduces total data volume, improving transmission efficiency without compromising diagnostic accuracy in critical areas.
Solution Approach 2:
The patent applies high quality only locally to the ROI where diagnostic accuracy is essential, while using low quality for the background region. This localized quality distribution optimizes the balance between diagnostic needs and transmission efficiency.
3Reliability
If conventional anonymization methods are applied, then patient confidentiality is protected, but crucial contextual data is lost
Solution Approach 1:
The patent applies anonymization selectively to the background region through downsampling, which removes identifying features while preserving the ROI at high resolution. This local approach protects patient confidentiality by obscuring non-critical areas while maintaining all crucial contextual data within the ROI for research analysis.
Solution Approach 2:
By segmenting the image into ROI and background, the patent enables differential processing: the background is downsampled for anonymization while the ROI retains full quality and contextual information. This segmentation allows confidentiality protection without sacrificing essential research data.
4Quantity of substance
If the background region is removed entirely, then file size is reduced, but spatial context information is lost
Solution Approach 1:
Instead of removing the background entirely, the patent applies low resolution downsampling to the background region. This approach maintains the spatial context and anatomical relationships necessary for understanding the ROI's location and orientation, while significantly reducing file size compared to full-resolution background retention.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods 300 and apparatus for recording anonymized volumetric data from medical image visualization software are disclosed. An example method includes receiving a medical image; anonymizing 312 the medical image outside of a region of interest by degrading the medical image outside a region corresponding to the region of interest and maintaining the resolution of the medical image inside the region corresponding to the region of interest; extracting data 402 from the medical image; and generating 406 an archive based on the anonymized medical image and the data.