About this video

• Video Title: Webinar: Advanced Acceleration Technology in MRI - An Overview
• Channel: IPEM Conferences
• Speakers: Stephen Jackson, Pauline (host)
• Duration: 1:00:34

Overview

This webinar provides an overview of Advanced Acceleration Technology (AAT) in MRI, focusing on compressed sensing, simultaneous multi-slice (SMS), and AI-based techniques. It discusses the underlying science, implementation guidance, image quality considerations, and shares user experiences and case studies from various manufacturers.

Key takeaways

• Benefits of AAT: Accelerated MRI sequences can lead to shorter patient appointments, reduced discomfort, increased patient throughput for radiology managers, quicker reporting, fewer motion artifacts for radiologists, and improved radiographer efficiency.
• Core AAT Techniques: The webinar focuses on Compressed Sensing (CS), Simultaneous Multi-Slice (SMS), and AI-based reconstructions, explaining their principles and how they differ from traditional acceleration methods like partial Fourier and parallel imaging.
• Compressed Sensing (CS): This technique involves incoherent subsampling of k-space, transformation into a sparse domain, and iterative reconstruction to denoise images, allowing for significant acceleration without a traditional SNR penalty.
• Simultaneous Multi-Slice (SMS): SMS uses specialized RF pulses to excite multiple slices simultaneously, which are then separated using unwrapping algorithms, effectively acquiring more data in the same amount of time.
• AI in MRI: AI techniques, including deep learning and neural networks, are used as an additional image processing step to improve image quality (SNR, sharpness) or reduce acquisition time by reconstructing higher-quality images from lower-SNR or lower-resolution acquisitions.
• Manufacturer Specifics & Implementation: The presentation highlights that AAT implementations are manufacturer-specific (GE, Siemens, Philips), with varying availability, trade names, and applications. Successful deployment requires careful testing, protocol optimization, and consideration of image quality and potential artifacts.
• User Experience & Case Studies: Real-world examples demonstrate significant time savings and comparable or improved image quality across different manufacturers and anatomies. However, reconstruction times can be longer, and unique artifacts may be encountered if the technology is pushed too far.
• Task and Finish Group: The IPEM task and finish group aims to raise awareness, share knowledge, and provide resources on AAT implementation through websites, webinars, and reports.

Accelerated MRI sequences offer several potential benefits across different groups within a radiology department:

• For Patients:

  • Shorter appointment times.
  • Less discomfort during the scan.
  • Reduced hesitancy to attend appointments due to shorter scan durations.

• For Radiology Managers:

  • Increased patient throughput.
  • Ability to meet national aims for increased diagnostic levels.

• For Radiologists:

  • Potential for quicker reporting due to better patient compliance.
  • Fewer motion artifacts, leading to clearer images.

• For Radiographers:

  • Improved patient compliance.
  • Shorter scan times, leading to less time owing (accrued downtime or delays).

The fundamental difference lies in how they achieve acceleration and reconstruct the image:

• Traditional Acceleration (e.g., Parallel Imaging): This method works by acquiring fewer lines of k-space (data collected during an MRI scan) than a fully sampled scan. To reconstruct the image, it relies on information about the sensitivities of the different receive coil elements used in the scanner. While it speeds up acquisition, it typically leads to a compromise in signal-to-noise ratio (SNR) or spatial resolution.

• Compressed Sensing (CS): CS also involves acquiring a subset of k-space data (sub-sampling). However, instead of relying solely on coil sensitivity information, CS leverages the fact that many MR images are "sparse" in certain mathematical domains (like wavelets). It uses an incoherent sub-sampling pattern in k-space, which, when reconstructed using iterative algorithms, allows for significant denoising and the recovery of image quality. This means CS can achieve higher acceleration factors than traditional methods without the usual penalty in SNR or spatial resolution, and it specifically aims to reduce noise and artifacts introduced by the sub-sampling.

AI-based reconstruction techniques aim to improve MRI scans in several ways:

• Boosting SNR: They can enhance the signal-to-noise ratio, allowing for faster imaging while maintaining image quality.
• Increasing Image Sharpness: Some applications improve image sharpness, which can enable the use of smaller matrix sizes for the same image quality, further contributing to faster scans or improved SNR.
• Reducing Artifacts: Certain AI techniques are designed to intrinsically reduce characteristic MRI artifacts.

A key characteristic of their clinical deployment is that the software is not learning in real-time. Once the neural network or AI algorithm is trained by the manufacturer and its parameters are set, this forms the fixed clinical product. The software does not continuously improve or adapt based on the images it processes in the clinic; it performs as it was designed and validated.

AI-based reconstruction techniques aim to improve MRI scans in several ways:

• Boosting SNR: They can enhance the signal-to-noise ratio, allowing for faster imaging while maintaining image quality.
• Increasing Image Sharpness: Some applications improve image sharpness, which can enable the use of smaller matrix sizes for the same image quality, further contributing to faster scans or improved SNR.
• Reducing Artifacts: Certain AI techniques are designed to intrinsically reduce characteristic MRI artifacts.

A key characteristic of their clinical deployment is that the software is not learning in real-time. Once the neural network or AI algorithm is trained by the manufacturer and its parameters are set, this forms the fixed clinical product. The software does not continuously improve or adapt based on the images it processes in the clinic; it performs as it was designed and validated.

The implementation of Simultaneous Multi-Slice (SMS) and Compressed Sensing (CS) technologies can present certain challenges and potential artifacts:

Challenges:

• Reconstruction Times: While the acquisition is faster, the reconstruction process for both SMS and CS can take longer than traditional methods. This is because iterative algorithms are often involved, especially with CS.
• Manufacturer-Specific Nuances: Each manufacturer has its own terminology, implementation details, and software packages for these technologies. This can create a learning curve and require specific training for each system.
• Availability and Cost: The availability of these technologies can vary significantly depending on the scanner model, software version, and hardware. Determining the cost and payback time for implementation can also be complex.
• Protocol Optimization: Implementing AAT across an entire protocol base requires significant work. It's not always realistic to implement it everywhere at once, and careful assessment with radiologists is recommended.
• SAR Concerns (SMS): For Siemens implementations of SMS, the use of novel RF pulses that excite multiple slices simultaneously may require more careful optimization for Specific Absorption Rate (SAR) to avoid hotspots, especially near implants.

Potential Artifacts:

• Exaggerated Motion Artifacts (CS): Motion artifacts, when they occur in CS sequences, tend to be more pronounced compared to non-CS sequences.
• Bright Fringe Artifacts (SMS): If the SMS factor is pushed too high, bright fringe artifacts can appear.
• Parallel Line Artifacts (CS): Characteristic parallel line artifacts can arise from compressed sensing reconstructions, particularly in situations with strong patient motion.
• Sensory Reconstruction Artifacts (AI/Deep Resolve Boost): In Siemens' Deep Resolve Boost, artifacts can occur as a relic of issues in the reconstruction part of the algorithm.
• General Artifacts with High Acceleration: Pushing any AAT technology to very high acceleration factors can lead to novel or exaggerated artifacts. Testing each deployment is crucial to ensure acceptable image quality and to identify and mitigate any new artifacts.