Anonymizing Medical Dose Report Images via OCR Signal Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for anonymizing dose report images in medical imaging examinations using ionizing radiation face challenges in accurately removing personal data due to variability in character sets and fonts, and existing solutions do not effectively ensure the complete absence of personal data in anonymized images.
Innovation Solution
A method involving optical character recognition to identify and anonymize specific image areas containing personal data within dose report images, using signal words to determine the positions of personal and dose data, and replacing these areas with constant or random values to ensure complete anonymization, preventing data extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If optical character recognition is used to identify and anonymize personal data in dose report images, then the reliability of anonymization is improved, but the device complexity and processing time increase due to the need to handle variable character sets and fonts
Solution Approach 1:
The patent extracts and removes personal data from dose report images by identifying specific regions containing personal information and replacing them with anonymized placeholders. This extraction approach ensures complete removal of personal data while maintaining the overall image structure and dose information integrity.
Solution Approach 2:
The system performs preliminary classification of image regions to identify whether areas contain personal data or dose information before applying anonymization. This preliminary action allows the system to selectively process only relevant regions, improving efficiency and reducing unnecessary processing complexity.
2Reliability
If complete anonymization of image areas is performed to ensure no personal data remains, then data protection compliance is improved, but information loss occurs in the anonymized regions
Solution Approach 1:
The patent carefully extracts only the personal data portions from the dose report images while preserving the rest of the image content. By precisely delimiting anonymization regions based on detected personal data boundaries, the system removes only necessary information while maintaining all dose-related information intact.
3Adaptability or versatility
If signal words are used to determine positions of personal and dose data, then the adaptability to different image formats is improved, but the difficulty of detecting and measuring increases due to character variability
Solution Approach 1:
The patent uses signal word patterns as templates to identify and locate personal data regions in various image formats. By copying and matching known signal word structures, the system can adapt to different formats and layouts without requiring complete retraining, effectively handling character set and font variations.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach reliably anonymizes dose report images, ensuring compliance with data protection regulations by effectively removing personal data and preventing its transmission to third-party providers, even when image formats change.
Implementation Method 1
a first position of the first signal word in the personal dose report image is determined by optical character recognition via a first computing unit
Data Source
AI summary
In a method, a personal dose report image is received via a first interface. The personal dose report image includes personal data and dose data of an imaging by way of ionizing radiation. The personal data and the dose data are contained as rasterized text in the personal dose report image and the personal data or the dose data includes a first signal word. A first position of the first signal word in the personal dose report image is determined through optical character recognition. An image area in the personal dose report image is determined based on the first position, the image area including at least a part of the personal data. An anonymized dose report image based on the personal dose report image is determined, at least the image area of the anonymized dose report image corresponding to the image area of the personal dose report image being anonymized.