Multi-B-Value DWI Lesion Detection With Transformer Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lesion detection and segmentation in diffusion-weighted imaging (DWI) using single b-value images is inadequate as it either provides clear images without identifiable lesions or clearly identifies lesions but with reduced clarity, failing to leverage the benefits of both low and high b-values effectively.
Innovation Solution
A multi-b value DWI image processing mechanism utilizing a multi-head transformer that combines lesion detection and segmentation, leveraging both low and high b-value images through a combined loss function, extracting inter-b-value features, and generating lesion masks using a multi-head transformer architecture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If single b-value DWI images are used for lesion detection, then image clarity is maintained, but lesion identification capability deteriorates
Solution Approach 1:
The patent combines multiple DWI images with different b-values into a unified processing framework. The system ingests a plurality of DWI images acquired at different b-values and processes them together through a neural network to simultaneously preserve image clarity and enhance lesion identification, resolving the contradiction between maintaining clarity and improving lesion detectability
2Reliability
If high b-value DWI images are used for lesion detection, then lesion identification capability is improved, but image clarity deteriorates
Solution Approach 1:
The patent applies different processing strategies to different regions and features within the DWI images. The neural network learns to extract and weigh different characteristics from various b-value images locally, preserving clarity in regions where it matters while enhancing lesion contrast in regions where lesions are present, thus resolving the global contradiction through local optimization
3Device complexity
If multiple b-value DWI images are processed separately, then processing simplicity is maintained, but feature extraction completeness deteriorates
Solution Approach 1:
The patent merges the processing of multiple b-value images into a single integrated neural network pipeline. Instead of processing images separately and combining results, the system processes all b-value images simultaneously through shared and specialized network layers, extracting comprehensive features while maintaining processing efficiency and avoiding information loss
Data Source
AI summary
Mechanisms are provided for detecting lesions in diffusion weighted imaging (DWI) images. The mechanisms receive a first set of DWI images corresponding to a anatomical structure, from medical imaging computer system(s). The first set of DWI images comprises a plurality of DWI images having at least two different b-values. The mechanisms generate a second set of DWI images from the first set of DWI images based on at least one predetermined criterion. The second set of DWI images comprises different DWI images having different b-values. The mechanisms extract feature data from the second set of DWI images, input the feature data into at least one computer neural network, and generate an output from the neural network(s) comprising at least one of a lesion classification or a lesion mask based on results of processing, by the neural network(s), of the feature data extracted from the second set of DWI images.