Medical Image Acquisition Metadata from Image and Non-Image Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automation methods for medical image processing in radiology workflows are inadequate due to insufficiently documented image acquisition parameters, leading to inappropriate results and inefficient resource consumption.
Innovation Solution
A computer-implemented method using trained functions to determine image acquisition information from both image and non-image data, including natural language processing and convolutional neural networks, to accurately extract relevant parameters for automated processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image acquisition information is read from documentation, then automation of medical image processing is enabled, but accuracy deteriorates due to insufficient or incorrect documentation
Solution Approach 1:
The patent introduces trained functions (machine learning models) as intermediaries between the medical image data and the automation workflow. These trained functions analyze the image data directly to extract image acquisition information, acting as a mediator that bridges the gap between incomplete documentation and accurate processing requirements. The trained functions process both image data and non-image data to derive reliable acquisition parameters without depending on potentially erroneous manual documentation.
2Productivity
If automation is based on documented information, then processing speed is improved, but reliability deteriorates due to dummy entries and incorrect protocols
Solution Approach 1:
The system enables self-service by allowing the medical image data itself to provide the acquisition information through automated analysis by trained functions. Instead of relying on external documentation that may contain errors, the image data undergoes self-analysis to extract its own acquisition parameters. This self-service approach ensures that the information used for automation is directly derived from the actual image characteristics rather than from potentially incorrect external sources.
Solution Approach 2:
The patent implements feedback mechanisms where the trained functions continuously learn from and adapt to the actual image data characteristics. The system uses feedback loops to refine the extraction of image acquisition information, comparing predicted values with actual image properties and adjusting accordingly. This feedback ensures that even when initial documentation is incorrect, the system can correct itself and maintain high reliability in automation decisions.
3Adaptability or versatility
If multiple imaging modalities and protocols are supported, then versatility is improved, but device complexity increases
Solution Approach 1:
The patent applies universality by designing trained functions that can handle multiple imaging modalities (CT, MRI, X-ray, etc.) and various protocols through a single unified framework. The machine learning models are trained on diverse datasets encompassing different modalities and protocols, enabling them to universally extract acquisition information across various imaging types. This multi-functional approach allows the system to support numerous imaging scenarios without requiring separate specialized processing paths for each modality.
Data Source
AI summary
Computer-implemented methods and systems for providing an image acquisition information of a medical image are provided. A method includes obtaining the medical image; extracting image data from the medical image; obtaining non-image data associated with the image data of the medical image; providing a first trained function configured to determine an image acquisition information based on the non-image data; providing a second trained function configured to determine an image acquisition information based on the image data of the medical image; determining a first image acquisition information by applying the first trained function to the non-image data; determining a second image acquisition information by applying the second trained function to the image data; determining the image acquisition information based on the first image acquisition information and the second image acquisition information; and providing the image acquisition information.