CT Image Style Transformation for Low-Dose Subtraction Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional CT subtraction procedures require an additional radiation dose to obtain a pre-contrast volume, which is not aligned with the post-contrast volume, leading to poor subtraction results and missing diagnostic signals.
Innovation Solution
A method and apparatus that utilize a 3D scanogram, acquired with a lower radiation dose, to transform its image style to resemble a conventional pre-contrast volume, enabling its use in subtraction procedures without additional scanning, by adjusting field of view, resolution, and intensity profiles using an adapter model trained on anatomical structure and intensity histograms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a conventional pre-contrast volume is obtained using CT imaging, then the subtraction procedure can be performed, but an additional radiation dose is required
Solution Approach 1:
The patent generates a synthetic pre-contrast volume by transforming the scanogram image data through a neural network adapter, creating a copy that resembles a conventional pre-contrast volume without requiring actual additional scanning. This synthetic copy is then used in the subtraction procedure, eliminating the need for an additional radiation dose while maintaining subtraction result quality
Solution Approach 2:
The patent applies parameter changes by transforming the image style of the scanogram from its original low-dose characteristics to characteristics that resemble conventional pre-contrast volumes. The adapter model modifies parameters such as intensity profiles, contrast characteristics, and image quality attributes to match the target style, enabling the scanogram to serve dual purposes
2Object-affected harmful factors
If a scanogram is used as pre-contrast volume, then radiation dose is reduced, but image style transformation is required to match post-contrast volume characteristics
Solution Approach 1:
The patent introduces a neural network adapter as an intermediary component that bridges the scanogram and the desired pre-contrast volume style. This adapter is trained on paired data of scanograms and conventional pre-contrast volumes, automatically learning the transformation requirements and applying them to convert the scanogram image style without manual intervention
Solution Approach 2:
The patent replaces the mechanical system of additional physical scanning with a computational transformation system. Instead of performing another CT scan to obtain the pre-contrast volume, the system uses an adapter model to computationally transform the scanogram data, substituting physical measurement with information processing
3Area of stationary object
If the scanogram has different field of view and resolution from the post-contrast volume, then the scanogram can cover larger areas, but alignment and subtraction quality deteriorate
Solution Approach 1:
The adapter model applies parameter changes to transform the scanogram's imaging parameters including field of view, resolution, and intensity characteristics. By adjusting these parameters during the style transformation, the scanogram is converted to match the spatial and quality characteristics of the post-contrast volume, enabling precise alignment and high-quality subtraction while maintaining the original scanogram's comprehensive coverage
Data Source
AI summary
A medical image processing apparatus comprises processing apparatus configured to: receive CT (computed tomography) image data acquired using a first scan setting; input the CT image data to an adapter to obtain corrected CT image data, wherein the adapter is configured to transform a first image style of the CT image data acquired using the first scan setting to a second, different image style that is characteristic of CT image data acquired using a second scan setting that is different from the first scan setting; and output the corrected CT image data.