Scout Image Anatomical Range Estimation Using CNNs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for setting the imaging range in medical imaging, such as CT and MRI, rely on manual operator input from scout images, which are of lower resolution and have larger slice intervals, leading to inaccuracies and incomplete inclusion of anatomical structures in the final three-dimensional images.
Innovation Solution
A derivation model is constructed using convolutional neural networks (CNN) trained with high-resolution and low-resolution three-dimensional images to accurately estimate both easily and hard-to-recognize ranges of anatomical structures in scout images, allowing for precise imaging range setting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual setting of imaging range is performed by operator viewing scout image, then flexibility and adaptability are maintained, but time consumption increases and setting accuracy varies
Solution Approach 1:
The system performs automatic imaging range setting without requiring manual operator intervention. The deep learning model processes scout images autonomously to determine imaging ranges, eliminating the need for operators to manually measure and set parameters while maintaining high accuracy consistent with expert human performance
Solution Approach 2:
The manual mechanical process of visual inspection and manual setting by operators is replaced with an automated computational system. The deep learning model uses neural networks to automatically detect anatomical structures and determine imaging ranges from scout images, substituting human cognitive and manual operations with algorithmic processing
2Productivity
If imaging range is set based only on scout image, then setting process is simple and fast, but required anatomical structure may not be included in three-dimensional image
Solution Approach 1:
The deep learning model acts as an intermediary between the scout image and the final imaging range determination. It processes the low-resolution scout image through neural network layers to infer the complete anatomical structure range, serving as a computational mediator that bridges the resolution gap between scout images and required three-dimensional imaging coverage
Solution Approach 2:
The system performs preliminary processing of the scout image through deep learning analysis before final imaging range setting. The model pre-processes the scout image to identify and expand the imaging range to ensure complete anatomical coverage, performing the complex analysis task before the actual imaging procedure
3Measurement precision
If deep learning model is trained with multiple three-dimensional images with different slice intervals, then estimation accuracy of anatomical structure range is improved, but training data requirements and model complexity increase
Solution Approach 1:
The training data is systematically varied by changing the slice interval parameter across multiple three-dimensional images. The deep learning model is trained on images with different slice intervals (e.g., 5mm, 10mm, 15mm) to learn invariant features that generalize across different imaging resolutions, improving robustness without requiring excessive training data
Solution Approach 2:
The training process is segmented into structured phases: collecting three-dimensional images with different slice intervals, preparing corresponding anatomical structure range annotations, dividing data into training/validation sets, and performing staged model training. This segmentation of the complex training process makes it manageable and systematic
Data Source
AI summary
An image processing device includes a processor, in which the processor is configured to: derive a first range of an anatomical structure, which is relatively easy to visually recognize in at least one tomographic image including the anatomical structure, and a second range of the anatomical structure, which is relatively hard to visually recognize in the at least one tomographic image, from the at least one tomographic image by using a derivation model; and display the first range and the second range.