Computer-Aided Diagnosis Feature Selection for Lesion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer-aided diagnosis (CAD) systems face challenges in accurately segmenting medical images and training classifiers to detect lesions, leading to inefficiencies and inaccuracies in diagnosis, particularly for pulmonary nodules and other medical conditions.
Innovation Solution
The method involves acquiring medical images, segmenting them using multiple models such as morphological and statistical models, extracting features, selecting a reference subset of features using algorithms like simulated annealing, and training classifiers to improve image analysis and diagnosis accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple segmentation models are used to segment medical images, then segmentation accuracy is improved, but system complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into multiple stages: preliminary positioning to identify candidate regions, threshold segmentation to create initial masks, and refined segmentation using statistical models to finalize lesion boundaries. This multi-stage segmentation approach improves accuracy while managing complexity through hierarchical decomposition of the segmentation process.
Solution Approach 2:
The patent implements a universal feature extraction framework that works across multiple segmentation models and lesion types. The same feature extraction and classification pipeline processes results from morphological models, statistical models, and hybrid approaches, reducing overall system complexity through standardized multi-functional processing components.
2Measurement precision
If a large number of features are extracted from segmented images, then classification accuracy improves, but training time and computational resources increase
Solution Approach 1:
The patent extracts only the most relevant features from segmented images for classification. Instead of using all possible image features, the system selectively extracts morphological features (area, perimeter, circularity), texture features, and intensity features that are most discriminative for lesion characterization, reducing training time while maintaining accuracy.
Solution Approach 2:
The patent applies local quality by differentiating feature extraction based on lesion type and location. Different feature sets are extracted depending on the specific clinical question and lesion characteristics, optimizing the balance between accuracy and computational efficiency for each specific diagnostic task rather than using a uniform comprehensive feature set.
3Reliability
If multiple segmentation models are employed, then detection reliability improves, but processing time increases
Solution Approach 1:
The patent performs preliminary positioning to identify candidate lesion regions before applying multiple segmentation models. This preliminary step creates initial masks that constrain subsequent segmentation operations, allowing multiple models to process only relevant regions rather than entire images, thereby maintaining reliability while reducing overall processing time.
Solution Approach 2:
The patent applies partial action by using multiple segmentation models selectively rather than uniformly across all images. The system chooses appropriate segmentation approaches based on lesion characteristics and clinical requirements, applying more models for complex cases and fewer for straightforward cases, optimizing the reliability-speed tradeoff.
Data Source
AI summary
The present disclosure relates to a method for training a classifier. The method includes: acquiring an original image; determining a candidate target by segmenting the original image based on at least two segmentation models; determining a universal set of features by extracting features from the candidate target; determining a reference subset of features by selecting features from the universal set of features; and determining a classifier by performing classifier training based on the reference subset of features.