Malignant Mass Detection in Radiographic Images Using Curvature Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer-aided detection (CAD) systems for radiographic images face challenges in accurately identifying malignant masses, particularly due to factors like occlusion, image complexity, and radiologist fatigue, which can lead to difficulties in detecting faint or subtle lesions.
Innovation Solution
A CAD system that includes a segmentation unit, detection units for calcifications, density, and distortions, and a classifier to analyze mammographic images, using a novel approach to estimate probability density functions (PDFs) for complex multidimensional distributions, enabling the detection and classification of potentially malignant masses by processing digital images and compensating for intensity gradients and tissue thickness variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If visual inspection by radiologists is used to detect masses, then the system is simple and fast, but detection accuracy decreases for faint or occluded masses
Solution Approach 1:
A computer-aided detection (CAD) system serves as an intermediary between the radiographic image and the radiologist, automatically analyzing images to detect masses and presenting results to the radiologist for final interpretation, thereby improving detection accuracy while maintaining workflow efficiency
Solution Approach 2:
The patent replaces the purely human visual inspection mechanism with a computer-based image analysis system that uses automated algorithms to detect mass signatures, substituting mechanical/human processes with computational methods to enhance reliability
2Reliability
If CAD algorithms are implemented to assist radiologists, then detection accuracy for faint lesions improves, but false positive rates increase
Solution Approach 1:
The CAD system dynamically adjusts detection parameters and thresholds based on image characteristics and mass signatures, optimizing sensitivity and specificity to improve detection accuracy while minimizing false positives through adaptive parameter tuning
Solution Approach 2:
The system incorporates feedback mechanisms where detection results are evaluated and used to refine subsequent detections, allowing the CAD system to learn from false positives and improve its classification accuracy over time
3Measurement precision
If traditional PDF approaches are used for classification, then the method is simple, but accuracy decreases for complex multidimensional distributions
Solution Approach 1:
The classification process is segmented into multiple stages: feature extraction, probability density function estimation for each feature, and combined classification decision, allowing complex multidimensional classification to be broken down into manageable components that improve accuracy
Solution Approach 2:
The patent extends traditional one-dimensional PDF approaches to multidimensional space by estimating joint probability density functions across multiple features simultaneously, adding dimensional complexity to capture complex mass signatures while improving classification precision
Data Source
AI summary
An image analysis embodiment comprises subsampling a digital image by a subsample factor related to a first anomaly size scale, thereby generating a subsampled image, smoothing the subsampled image to generate a smoothed image, determining a minimum negative second derivative for each pixel in the smoothed image, determining each pixel having a convex down curvature based on a negative minimum negative second derivative value for the respective pixel, joining each eight-neighbor connected pixels having convex down curvature to identify each initial anomaly area, selecting the initial anomaly areas having strongest convex down curvatures based on a respective maximum negative second derivative for each of the initial anomaly areas, extracting one or more classification features for each selected anomaly area, and classifying the selected anomaly areas based on the extracted one or more classification features.