Multi-sequence MR image-based synthetic CT generation using a generative adversarial network for head and neck
Mengke Qi1, Yongbao Li2, Aiqian Wu1
1Department of Biomedical Engineering, Southern Medical University, Guangzhou, 510515, Guangdong, China.
Medical Physics
|February 7, 2020
Summary
Multi-channel magnetic resonance (MR) sequences improve deep learning-based synthetic CT (sCT) accuracy for head and neck imaging. T1-weighted MR images alone offer sufficient data for sCT generation in clinical settings.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Deep learning models are increasingly used for generating synthetic computed tomography (sCT) images from magnetic resonance (MR) data.
- Accurate sCT generation is crucial for radiation therapy planning, particularly in complex anatomical regions like the head and neck.
- The choice of MR sequences as input significantly impacts the performance of these deep learning models.
Purpose of the Study:
- To evaluate the impact of different magnetic resonance (MR) imaging sequences on the accuracy of deep learning-based synthetic computed tomography (sCT) generation.
- To compare the performance of conditional generative adversarial networks (cGANs) using single and multi-channel MR inputs for sCT synthesis.
- To assess the suitability of T1-weighted MR images for sCT generation in clinical scenarios.
Main Methods:
- Four MR sequences (T1, T2, T1C, T1DixonC-water) were acquired from 45 nasopharyngeal carcinoma patients.
- Seven conditional generative adversarial network (cGAN) models were trained using single and combined MR sequences.
- A U-net network was used as a comparative model; performance was evaluated using metrics like Mean Absolute Error and Dice Similarity Coefficient.
Main Results:
- The cGAN model utilizing all four MR sequences (multi-channel) demonstrated superior accuracy compared to single-sequence models.
- Among single sequences, T1-weighted MR images yielded the best sCT prediction results.
- cGAN-generated sCTs exhibited higher fidelity, retaining more image details and reduced blurring compared to U-net predictions.
Conclusions:
- Conditional generative adversarial networks (cGANs) incorporating multiple MR sequences achieve the highest accuracy for synthetic CT (sCT) generation.
- T1-weighted MR images are a viable and effective input for sCT prediction, especially when acquisition time or sequence availability is limited.
- Deep learning-based sCT generation shows promise for clinical applications in head and neck imaging.
Keywords:
MRI-only radiotherapyconditional generative adversarial networkmulti-MR sequencesCT generationMore Related Videos
Related Concept Videos
01:24Magnetic Resonance Imaging
8.9K
Magnetic resonance imaging (MRI) is a noninvasive medical imaging technique based on a phenomenon of nuclear physics discovered in the 1930s, in which matter exposed to magnetic fields and radio waves was found to emit radio signals. In 1970, a physician and researcher named Raymond Damadian noticed that malignant (cancerous) tissue gave off different signals than normal body tissue. He applied for a patent for the first MRI scanning device in clinical use by the early 1980s. The early MRI...
8.9K
01:14Imaging Studies I: CT and MRI
718
Introduction: MRI and CT scans are crucial advancements in medical imaging techniques, playing a vital role in diagnosing conditions related to the gastrointestinal (GI) system. Each scan serves distinct purposes, targets specific areas, and requires unique nursing duties.
Description of the Procedures
Computed Tomography (CT) scan:
Computed Tomography (CT) scans use X-ray technology to generate detailed images of bones, organs, and tissues. During the scan, the patient lies on a moving table...
Description of the Procedures
Computed Tomography (CT) scan:
Computed Tomography (CT) scans use X-ray technology to generate detailed images of bones, organs, and tissues. During the scan, the patient lies on a moving table...
718