Multi-phase Lesion Detection via Unified Feature Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting and characterizing lesions in multi-phase imaging are subjective, prone to errors, and do not effectively utilize dynamic enhancement patterns, leading to variability and uncertainty in clinical evaluations.
Innovation Solution
A deep learning method that involves acquiring multi-phase images, extracting local contexts, encoding these into phase-specific feature maps, and combining them to create a unified feature map for enhanced lesion detection and characterization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual evaluation by radiologists is used, then clinical experience can be applied, but subjectivity and variability increase
Solution Approach 1:
The patent replaces manual radiologist evaluation with an automated deep learning system that processes multi-phase imaging data. The system uses neural networks to automatically detect and characterize lesions, eliminating human subjectivity and variability while maintaining clinical expertise through trained algorithms.
Solution Approach 2:
The deep learning system performs self-evaluation by automatically analyzing multi-phase images, extracting features, and generating lesion characterizations without requiring manual radiologist intervention for each case, thereby improving consistency and reducing variability in clinical evaluations.
2Reliability
If deep learning methods work on single phase images or in a phase-agnostic manner, then processing is simpler, but dynamic enhancement patterns are ignored
Solution Approach 1:
The patent segments the multi-phase imaging data into phase-specific feature maps, where each phase is processed independently to capture its unique characteristics. This segmentation allows the system to preserve dynamic enhancement patterns while managing complexity through modular phase-specific processing followed by fusion.
Solution Approach 2:
The patent combines phase-specific feature maps into a unified representation that integrates information from all phases. This merging process enables the system to leverage dynamic enhancement patterns across phases while maintaining the benefits of phase-specific analysis, thereby improving lesion characterization accuracy.
3Reliability
If 3D models are used to take advantage of volumetric context, then structural information is improved, but computational complexity increases
Solution Approach 1:
The patent segments volumetric data into phase-specific 2D feature maps rather than processing the entire 3D volume at once. This segmentation reduces computational complexity by breaking down the large-scale 3D problem into manageable phase-specific components while preserving essential volumetric context through the multi-phase approach.
Solution Approach 2:
The patent extracts phase-specific features from volumetric data, isolating the most relevant information from each phase. This extraction process reduces the computational burden by focusing on key features rather than processing all volumetric data, thereby managing model complexity while maintaining diagnostic accuracy.
4Ease of operation
If detection and characterization are performed on individual phases separately, then processing is more manageable, but performance degrades when phase acquisition is not standardized
Solution Approach 1:
The patent merges phase-specific feature maps into a unified representation that integrates information across all phases. This fusion approach makes the system robust to variations in phase acquisition protocols because the combined representation leverages complementary information from multiple phases, improving reliability under non-standardized conditions while maintaining processing manageability.
Data Source
AI summary
A method for detection and characterization of lesions includes acquiring a plurality of phase images of a multi-phase imaging exam, extracting a local context for each phase image of the plurality of phase images, encoding the local contexts to create phase specific feature maps, combining the phase-specific feature maps to create unified feature maps, and at least one of characterizing or detecting a lesion based on the unified feature maps.