Deep Neural Network Diagnostic Image Translation Using Parameter Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-based radiological imaging models typically ignore imaging parameters, limiting their accuracy in image translation tasks, especially when flexible imaging parameters are used.
Innovation Solution
Incorporating imaging parameters as additional input to a deep neural network, specifically through parameter image maps at every pixel, to improve prediction accuracy in diagnostic imaging tasks such as T1-weighted image translation and water-fat separation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models are trained to learn physical models from radiological images alone, then the model structure remains simple and easy to implement, but the prediction accuracy deteriorates when flexible imaging parameters are used
Solution Approach 1:
The patent merges radiological images and imaging parameter maps as combined inputs to the deep neural network. The input layer receives both the radiological image data and the corresponding imaging parameter values organized as parameter image maps, creating a unified input structure that leverages both visual information and quantitative parameter data to improve prediction accuracy in image translation tasks
2Measurement precision
If imaging parameters are incorporated as additional input to the deep neural network, then prediction accuracy improves, but the device complexity and computational requirements increase
Solution Approach 1:
The patent transforms imaging parameters into parameter image maps that match the spatial dimension of the radiological images. By organizing parameter values as 2D or 3D arrays with the same dimensions as the input images, the system adds a new dimensional aspect to the data representation, enabling the network to process both image intensity and parameter information simultaneously through convolutional operations
Data Source
AI summary
A deep learning method generates translated images from both diagnostic images acquired with predetermined image acquisition parameters and from the predetermined image acquisition parameters. The translated diagnostic images are generated by applying both the predetermined image acquisition parameters and the diagnostic images as input to the deep neural network in the form of parameter image maps with imaging parameter values at each pixel of the parameter image maps.