Medical Image Feature Localization Using Descriptor Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image registration and landmark detection techniques for medical imaging are computationally demanding and limited in accuracy and flexibility, especially when images are captured with different modalities or protocols, making it difficult to accurately locate a given feature across multiple medical images.
Innovation Solution
A method that determines the location of a given feature in medical imaging data by calculating similarity metrics between predefined descriptors for known and candidate locations, allowing for efficient and flexible feature location without relying on voxel-to-voxel mapping or trained classifiers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image registration techniques with voxel-to-voxel mapping are used, then location accuracy can be improved, but computational demand increases significantly
Solution Approach 1:
The patent extracts and compares only the essential descriptor features from each image region rather than performing comprehensive voxel-to-voxel mapping. By extracting descriptors that capture key visual characteristics and comparing only these extracted features, the system achieves accurate location identification without the computational burden of full image registration.
Solution Approach 2:
The patent segments the image into multiple regions and extracts descriptors for each region independently. This segmentation approach allows the system to process and compare only the relevant visual features in each segment rather than the entire image, reducing computational complexity while maintaining location accuracy.
2Measurement precision
If landmark detection with trained classifiers is used, then feature identification can be achieved, but training time and computational resources are consumed
Solution Approach 1:
Instead of training classifiers to identify landmarks, the patent uses a copying approach where pre-computed image descriptors are directly compared to identify matching features. This eliminates the time-consuming training process while maintaining the ability to accurately identify and locate features across images.
Solution Approach 2:
The patent performs preliminary extraction of image descriptors in advance, creating a database of descriptors that can be quickly queried and matched. This preliminary action avoids the need for time-consuming real-time training and classification, enabling rapid feature identification when needed.
3Measurement precision
If voxel-to-voxel mapping is used for image registration, then spatial alignment can be achieved, but the method is limited to similar imaging modalities
Solution Approach 1:
The patent employs a universal descriptor extraction and comparison method that can handle different imaging modalities (CT, MRI, PET, etc.) without requiring modality-specific processing. By extracting visual descriptors that capture essential image characteristics in a modality-independent manner, the system achieves spatial alignment across diverse imaging types.
Solution Approach 2:
The patent adapts to different imaging modalities by extracting descriptors that are robust to parameter variations between modalities. The descriptor extraction process is designed to capture invariant features that remain recognizable across different imaging parameters, enabling flexible application across multiple modalities.
4Productivity
If extensive pre-processing is performed to enable real-time results, then computational resources are reduced, but processing time increases
Solution Approach 1:
The patent extracts only the essential descriptor information from images in advance and stores it for rapid retrieval and comparison. By extracting and pre-storing only the necessary visual features rather than performing extensive pre-processing, the system enables real-time queries without significant pre-processing time costs.
Data Source
AI summary
A computer implemented method and apparatus for determining a location at which a given feature is represented in medical imaging data is disclosed. A first descriptor for a first location in first medical imaging data is obtained. The first location is the location within the first medical imaging data at which the given feature is represented. A second descriptor for each of a plurality of candidate second locations in second medical imaging data is obtained. A similarity metric indicating a degree of similarity with the first descriptor is calculated for each of the plurality of candidate second locations. A candidate second location is selected from among the plurality of candidate second locations based on the calculated similarity metrics. The location at which the given feature is represented in the second medical imaging data is determined based on the selected candidate second location.