Breast Mass Classification Using Spiculation and Density Measures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mammographic computer-aided detection (CAD) systems face challenges in accurately classifying mass-like regions in breast imagery, particularly those with both spiculated and dense characteristics, leading to high false positive rates and reduced sensitivity.
Innovation Solution
A method involving the computation of quantitative measures of spiculation and density for mass-like regions, followed by the selection and execution of appropriate classification schemes based on these measurements to differentiate between true positives and false positives, using independent detection algorithms for spiculation and density, and tailored classification algorithms for each characteristic type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single classification algorithm is used for all mass-like candidates, then the system is simple to operate, but classification accuracy is reduced for candidates with only spiculated or only dense characteristics
Solution Approach 1:
The classification system is segmented into multiple specialized algorithms: a spiculated mass classification algorithm for candidates with spiculated characteristics, a dense mass classification algorithm for candidates with dense characteristics, and a combined algorithm for candidates with both characteristics. Each algorithm is optimized for specific mass types, improving classification accuracy while maintaining manageable system complexity through modular design.
Solution Approach 2:
Different classification algorithms are applied to different regions of the candidate space based on their characteristic profiles. Candidates with predominantly spiculated characteristics receive spiculated-mass optimization, while those with dense characteristics receive dense-mass optimization. This localized approach ensures each candidate is evaluated by the most appropriate algorithm for its specific features.
2Reliability
If detection sensitivity is increased to identify all potential cancers, then true positive detection is improved, but false positive rate increases significantly
Solution Approach 1:
The system changes classification parameters based on the detected characteristics of each candidate. By computing spiculation and density measures and selecting algorithms accordingly, the system adapts its classification threshold and criteria to match the specific mass type, thereby maintaining high sensitivity for true positives while reducing false positives through characteristic-appropriate evaluation.
Solution Approach 2:
The classification system uses feedback from the computed spiculation and density measures to select the appropriate classification algorithm. This feedback mechanism ensures that candidates are evaluated using algorithms that have been optimized for their specific characteristics, improving overall detection reliability while minimizing false alarms.
Data Source
AI summary
Systems and methods are presented that detect and classify mass-like regions exhibiting spiculated and/or dense characteristics with high sensitivity and at acceptable false positive rates. One or more suspicious masses are identified in medical imagery of the breast. In certain embodiments, a quantitative measure of spiculation and quantitative measure of density are computed for each suspicious mass located. At least one classification scheme, developed using true and false positives with similar quantitative measures, is then selected for each suspicious mass according to both quantitative measures. In certain other embodiments, a measure of breast location is computed for each suspicious mass. In one embodiment, the location determines whether a suspicious mass appears inside or outside of the parenchyma region of the breast.