Lesion Segmentation Using Dual Segmenters to Reduce Manual Input
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Computer Aided Diagnosis (CAD) systems for lesion segmentation in medical images rely heavily on subjective user input and point-based methods, which are inefficient and prone to errors due to their dependence on manual control points and user criteria.
Innovation Solution
An apparatus and method that utilize two segmenters: a first segmenter generates a candidate lesion list based on examination information, and a second segmenter determines a target lesion for segmentation using adjusted parameters, including slope, region, and filtering terms, to produce a more accurate and objective segmentation result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a point-based segmentation method is used with manual control points, then the user can perform lesion segmentation, but the process requires heavy user input and is highly dependent on subjective user criteria
Solution Approach 1:
The system performs automatic lesion segmentation using a trained neural network model, eliminating the need for users to manually move control points. The CAD system segments lesions autonomously based on medical images and examination information, making the system self-sufficient and reducing user burden.
Solution Approach 2:
The patent replaces the mechanical point-based interaction system with an automated image processing system. Instead of users manually placing control points, the system uses a trained segmentation model that processes medical images automatically, substituting human manual operations with automated computational methods.
2Productivity
If manual point-based segmentation is used, then the user can segment lesions, but the method is inefficient and time-consuming
Solution Approach 1:
The system performs preliminary actions by pre-training a segmentation model using a dataset of medical images and examination information before actual use. This preliminary training enables the system to automatically segment lesions without requiring users to perform time-consuming manual operations during actual diagnosis.
Solution Approach 2:
The automated segmentation model performs lesion segmentation independently without requiring user intervention for each segmentation task, dramatically improving productivity and eliminating the time loss associated with manual point-based methods.
3Adaptability or versatility
If subjective user criteria are used for segmentation, then the user can apply personal judgment, but the results lack consistency and objectivity
Solution Approach 1:
The system uses examination information (such as examination type, body part, and clinical context) as input parameters to the segmentation model. These parameters guide the model to adapt its segmentation to different clinical scenarios while maintaining objective and consistent results across different users and cases.
Solution Approach 2:
The system uses a dataset comprising medical images, examination information, and ground truth lesion annotations to train the segmentation model. This feedback loop during training enables the model to learn optimal segmentation criteria that are both adaptable to different cases and consistent in their application.
Data Source
AI summary
An apparatus and method are provided including a first segmenter and a second segmenter. The first segmenter is configured to generate a first segmentation result from a medical image using a first segmentation parameter for a candidate lesion. The second segmenter is configured to determine a target lesion to segment from among the candidate lesion based on the first segmentation result, and generate a second segmentation result using a second segmentation parameter to segment the target lesion.