Mammographic Mass Identification via Multi-View Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mammographic lesion detection methods primarily focus on single-view analysis, failing to effectively capture and analyze the rich information from multi-view mammograms, which is crucial for accurate breast cancer diagnosis.
Innovation Solution
The proposed method and apparatus utilize a Faster-RCNN network module and DeepLab network module in parallel for simultaneous ipsilateral and bilateral analysis, explicitly encoding mass-to-nipple distance and incorporating symmetry constraints through a DNN-based nipple detector, ipsilateral analyzer, and bilateral analyzer to generate fused probability maps for improved lesion detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If single-view analysis is used for mammographic lesion detection, then the device complexity is reduced, but the measurement precision and detection accuracy deteriorate due to inability to capture rich multi-view information
Solution Approach 1:
The system segments the multi-view analysis into distinct functional modules: a nipple detector for locating nipple positions, an ipsilateral analyzer for same-breast view analysis, and a bilateral analyzer for opposite-breast view analysis. Each module processes specific aspects of the mammographic images independently, then their results are combined to achieve high-precision lesion detection without overwhelming system complexity
Solution Approach 2:
The system transitions from single-view (2D) analysis to multi-view (3D spatial information) analysis by incorporating both ipsilateral and bilateral mammographic views. This dimensional expansion provides depth information and different projection angles, significantly improving lesion detection accuracy while the modular architecture manages the resulting complexity
2Measurement precision
If multi-view analysis is implemented to capture rich information, then the measurement precision improves, but the device complexity and computational requirements increase
Solution Approach 1:
The complex multi-view analysis is divided into specialized sub-analyzers: the ipsilateral analyzer processes CC and MLO views of the same breast, while the bilateral analyzer processes corresponding views from the opposite breast. This segmentation allows each analyzer to focus on specific spatial relationships and symmetry patterns, improving mass identification accuracy while managing computational complexity through functional specialization
Solution Approach 2:
The nipple detector performs preliminary action by accurately locating nipple positions in both ipsilateral and bilateral views before the main analysis. These detected nipple locations serve as reference points that guide the subsequent ipsilateral and bilateral analyzers, pre-organizing the spatial framework and reducing the computational burden during the main lesion detection phase
3Productivity
If DNN-based approaches are used for mammogram analysis, then the productivity and detection performance improve, but the loss of information increases due to potential over-processing and false positives
Solution Approach 1:
The system implements feedback through cross-validation between ipsilateral and bilateral analysis. Lesion candidates detected in one view are verified against corresponding regions in opposite-breast views using symmetry constraints. This feedback mechanism reduces false positives by requiring consistent findings across multiple views while maintaining high detection performance through iterative refinement
Solution Approach 2:
The system applies partial analysis by focusing computational resources on regions of interest identified through nipple-detector-guided spatial relationships. Rather than processing entire images uniformly, the analyzers concentrate on specific anatomical regions where lesions are most likely to occur, improving productivity while reducing information loss through targeted rather than exhaustive processing
4Measurement precision
If existing DNN-based dual-view approaches are used, then some multi-view information is captured, but the measurement precision remains insufficient due to inadequate utilization of symmetry information and spatial relationships
Solution Approach 1:
The nipple detector performs preliminary localization of nipple positions in both ipsilateral and bilateral views, establishing a reference coordinate system before main analysis. This preliminary action enables the system to accurately compute spatial relationships and symmetry constraints, ensuring that no multi-view information is lost during subsequent processing while improving diagnostic precision through geometrically-aware analysis
Solution Approach 2:
The system adds the dimension of bilateral symmetry analysis by comparing corresponding regions between left and right breasts. This dimensional expansion from univariate to bivariate analysis captures spatial relationships and anatomical symmetries that single-view or simple dual-view approaches miss, improving measurement precision while preserving all multi-view information through comprehensive spatial modeling
Data Source
AI summary
A method, applied to an apparatus for mammographic multi-view mass identification, includes receiving a main image, a first auxiliary image, and a second auxiliary image. The main image and the first auxiliary image are images of a breast of a person, and the second auxiliary image is an image of another breast of the person. The method further includes detecting the nipple location based on the main image and the first auxiliary image; generating a first probability map of the main image based on the main image, the first auxiliary image, and the nipple location; generating a second probability map of the main image based on the main image, the second auxiliary image, and the nipple location; and generating and outputting a fused probability map based on the first probability map and the second probability map.