Synthetic CT Generation From MRI Using CNN Direct Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating synthetic CT images from MRI images are inefficient, requiring significant computing resources, time, and often result in inaccurate representations due to the lack of a direct mathematical relationship between CT and MRI intensity values, especially in distinguishing soft tissues like prostate and bladder.
Innovation Solution
A convolutional neural network (CNN) model is trained to convert MRI images to synthetic CT images, utilizing multi-channel MRI data and learning a direct mapping without the need for inter-subject image registration, allowing for faster and more accurate generation of synthetic CT images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If atlas-based methods are used to generate synthetic CT images from MRI images, then the synthetic images can be generated, but the computation time and computing resources required are significant and the accuracy is limited
Solution Approach 1:
The patent applies preliminary action by pre-training a convolutional neural network model on a large dataset of paired MRI and CT images before actual synthesis. This pre-training phase performs the computationally intensive learning of mapping relationships in advance, so that when synthetic CT images need to be generated from new MRI images, the pre-trained model can quickly perform the transformation without requiring extensive computation time during actual use.
Solution Approach 2:
The patent replaces the mechanical atlas-based registration system with a data-driven convolutional neural network approach. Instead of using traditional image registration algorithms that require iterative optimization and significant computational resources, the invention uses a trained neural network that has learned the complex non-linear mapping between MRI and CT intensity values, enabling faster and more accurate synthesis.
2Ease of manufacture
If tissue classification-based approaches are used to generate synthetic CT images, then the process can be simplified, but the accuracy in distinguishing soft tissues is poor
Solution Approach 1:
The patent applies parameter changes by transforming the input MRI image intensities through a learned non-linear mapping function to generate synthetic CT Hounsfield units. The convolutional neural network learns optimal intensity transformation parameters from training data, enabling accurate representation of different tissue types including soft tissues, without requiring manual tissue classification or segmentation.
3Measurement precision
If conventional CT imaging is used to obtain accurate patient geometry and electron density, then the radiation dose calculation is accurate, but the patient is exposed to additional radiation dosage
Solution Approach 1:
The patent introduces an intermediary approach by using MRI images as the primary input and generating synthetic CT images as an intermediate representation. The synthetic CT images serve as a mediator that provides the necessary electron density information for radiation dose calculation without requiring the patient to undergo actual CT scanning. This intermediary synthetic image allows accurate dose computation while avoiding additional ionizing radiation exposure.
Data Source
AI summary
Systems, computer-implemented methods, and computer readable media for generating a synthetic image of an anatomical portion based on an origin image of the anatomical portion acquired by an imaging device using a first imaging modality are disclosed. These systems may be configured to receive the origin image of the anatomical portion acquired by the imaging device using the first imaging modality, receive a convolutional neural network model trained for predicting the synthetic image based on the origin image, and convert the origin image to the synthetic image through the convolutional neural network model. The synthetic image may resemble an imaging of the anatomical portion using a second imaging modality differing from the first imaging modality.