Medical Image Metadata Routing via Neural Network Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Digital Imaging and Communications in Medicine (DICOM) format for medical images often results in incomplete or inaccurate metadata due to limitations in data entry and structure, particularly failing to support ambiguous data entry which compromises completeness, and existing solutions like the DICOM Imaging Router only support distinct anatomical groups and single-class images.
Innovation Solution
A method that performs object detection and classification on medical images to store detected parameters in a structured XML file or DICOM format, enabling the detection of multiple objects and classes within the same image, using a neural network for feature extraction and mapping, and a logic module for routing images to specialized processing modules based on detected classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard DICOM format is used for storing metadata, then data exchange compatibility is maintained, but metadata completeness and accuracy deteriorate due to structural limitations and lack of support for ambiguous data entry
Solution Approach 1:
The patent segments metadata storage into two parts: standard DICOM tags for basic compatibility and a separate XML file for extended structured metadata. This segmentation allows the system to maintain DICOM format compatibility while storing comprehensive multi-class object information that exceeds standard DICOM capabilities.
Solution Approach 2:
The patent introduces an XML file as an intermediary carrier for extended metadata. This intermediary structure bridges the gap between standard DICOM limitations and the need for comprehensive multi-class object information, allowing rich metadata storage without compromising DICOM compatibility.
2Productivity
If existing routing solutions like DICOM Imaging Router are used, then routing for distinct anatomical groups is achieved, but support for images with multiple overlapping classes deteriorates
Solution Approach 1:
The patent implements a routing system that universally handles multiple object classes within a single image. The neural network detects and classifies multiple overlapping objects (e.g., body parts, implants, outside structures) simultaneously, making the routing system adaptable to diverse image types beyond single-class limitations.
Solution Approach 2:
The patent uses a composite approach combining neural network-based multi-class detection with XML-based metadata storage and extended DICOM tags. This composite system integrates multiple components to achieve comprehensive multi-class object recognition and routing, overcoming the limitations of existing single-class solutions.
3Reliability
If manual data entry is used for metadata, then data accuracy can be maintained, but time consumption and completeness deteriorate
Solution Approach 1:
The patent implements automated object detection and classification using neural networks that process medical images and generate metadata automatically. This self-service approach eliminates manual data entry while maintaining high accuracy through AI-based detection, significantly reducing time consumption compared to manual methods.
Solution Approach 2:
The patent performs preliminary object detection and classification before final metadata generation. The neural network pre-processes images to identify multiple objects and their classes, preparing structured information that is then stored in XML and DICOM formats, streamlining the overall metadata generation process.
Data Source
Figure 1a
Figure 1b
Figure 1c
AI summary
A method for processing a medical image, the method comprising the following steps: receiving a medical image, performing an object detection and classification on said medical image, storing the detected parameters of one or more detected objects in association with the image.