Tomographic Image Positioning via Contrastive Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for manually setting imaging ranges in medical imaging require time and depend on operator skill, and scout images with larger slice intervals may not include target anatomical structures, making accurate range setting difficult.
Innovation Solution
A processor-based system using contrastive learning to derive normalized relative positions of tomographic images, enabling specification of target anatomical structures even when they are not directly visible in scout images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If scout imaging is performed with larger slice intervals to reduce imaging time and data volume, then productivity is improved, but measurement precision deteriorates because target anatomical structures may not be included in the tomographic images
Solution Approach 1:
The patent introduces a derivation model trained through contrastive learning as an intermediary between the scout image and the target anatomical structure position. This model learns the positional relationship between visible anatomical structures and target structures from training data, enabling it to predict target positions even when they are not directly visible in the scout image. The model acts as a mediator that bridges the gap between limited scout image data and accurate position specification requirements.
2Measurement precision
If manual setting of imaging range is performed by operators to ensure accurate positioning, then measurement precision is improved, but productivity deteriorates due to time consumption and dependence on operator skill
Solution Approach 1:
The system implements self-service by automatically determining the imaging range through the derivation model without requiring manual operator intervention. The model independently processes the scout image, predicts the position of target anatomical structures, and determines the optimal imaging range automatically. This eliminates the need for operators to manually adjust settings, thereby improving productivity while maintaining consistent accuracy regardless of operator skill level.
Solution Approach 2:
The patent replaces the mechanical system of manual operator judgment and adjustment with an automated computational system. Instead of relying on human visual inspection and manual range setting, the system uses the trained derivation model to automatically predict positions and determine imaging ranges. This substitution of manual mechanical operations with automated computational processing improves both efficiency and consistency.
3Measurement precision
If contrastive learning is applied to construct the derivation model for position prediction, then measurement precision is improved by enabling target specification without direct visibility, but device complexity increases due to the learning system requirements
Solution Approach 1:
The patent applies preliminary action by pre-training the derivation model using contrastive learning on a large dataset of tomographic images with known anatomical structure positions. This training phase is performed in advance before actual use, allowing the model to learn robust positional relationships and patterns. Once trained, the model can be deployed for rapid inference without requiring complex real-time processing, thereby reducing the complexity burden during operational use while maintaining high precision.
Data Source
AI summary
An image processing device includes a processor, in which the processor is configured to: input at least one processing target tomographic image to a derivation model constructed by contrastive learning using a plurality of tomographic images acquired by imaging an interior of a body such that a specific anatomical structure is included, the derivation model being constructed by the contrastive learning so as to derive a normalized relative position in the interior of the body based on a relative reference position, which is determined in advance for the specific anatomical structure, in the interior of the body; and derive a normalized relative position of the at least one processing target tomographic image in the interior of the body via the derivation model.