Deep Learning Positioning Model for Medical Scout Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual operation for determining a positioning frame in medical scanning is labor-intensive and requires high operator skill, limiting efficiency and accuracy.
Innovation Solution
A method and system that utilize a pre-trained positioning model, incorporating a segmentation model with encoding and decoding modules, to automatically determine a positioning frame from a scout image, based on gold standard positioning frames, improving efficiency and accuracy without subsequent processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual operation is used to determine positioning frame, then operator can control the positioning process, but the workload is great and requires high operator ability
Solution Approach 1:
The system enables self-service by allowing the positioning frame to be automatically determined through deep learning algorithms. The model independently processes scout images and generates positioning frames without requiring manual operator intervention, thereby eliminating the need for high operator ability while maintaining high positioning efficiency.
Solution Approach 2:
The patent replaces the manual mechanical operation system with an automated deep learning system. Instead of operators using tools like keyboards or mice to manually determine positioning frames, the system uses a trained deep learning model that automatically processes images and generates positioning frames, substituting human mechanical operations with automated computational processes.
2Device complexity
If manual operation mode is used for positioning frame determination, then flexibility in adjustment is maintained, but workload increases and operator ability requirement increases
Solution Approach 1:
The deep learning model performs self-service by automatically analyzing scout images and determining positioning frames without human intervention. This eliminates complex manual operations and reduces the operation process to a simple automated workflow, thereby reducing both device complexity and operator ability requirements.
Solution Approach 2:
The patent extracts the complex decision-making process from manual operations and embeds it within the trained deep learning model. The model has been pre-trained on gold standard positioning frames, extracting the expertise and decision-making logic into the model itself, thereby removing the need for operators to possess high ability requirements.
3Productivity
If automated positioning model is used, then workload is reduced and positioning speed is improved, but model training complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training the deep learning model offline using gold standard positioning frames before deployment. This preliminary training phase prepares the model in advance, allowing it to perform automated positioning efficiently during actual use without requiring complex real-time training, thereby achieving high productivity while managing training complexity through advance preparation.
Data Source
AI summary
The present disclosure discloses a method for positioning. The method may include obtaining a scout image of a target object and inputting the scout image of the target object into a positioning model. The method may also include determining a positioning frame in the scout image based on an output result of the positioning model. The positioning model may be obtained by training based on training scout images and information of the gold standard positioning frames. Each of the gold standard positioning frames may be in, from, or determined based on one of the training scout images and used for scanning.