It can enable more exams per day, shorten exam length, and improve image quality for challenging patients. These benefits have driven many industry players to innovate to increase speed and improve the quality of MRI.
With the use of deep-learning technologies, image reconstruction is now faster and yields significantly enhanced image quality. This development has expanded and broadened the capabilities of MRI.
Recognizing the potential of artificial intelligence (AI), Philips partnered with Leiden University Medical Center in the Netherlands to develop deep learning-based image reconstruction software that won the best AI reconstruction technology award in the fastMRI challenge organized by New York Langone Health and Facebook AI Research in 2019.
A recent addition to the portfolio, Philips MR SmartSpeed combines this award-winning AI technology with the state-of-the-art Compressed SENSE speed engine to deliver fast and high-quality imaging for the vast majority of patients. By combining fast data acquisition with AI reconstruction, SmartSpeed transforms MR imaging. This can improve radiologists' clinical confidence and increase MR department productivity, ultimately enhancing the overall quality of care for patients.
Enhancing diagnostic confidence
Thanks to improved image quality, deep learning-based reconstruction can help radiologists to identify abnormalities and have high diagnostic confidence in their interpretations. Many industry players have incorporated AI into their MR imaging solutions, but the impact of AI reconstruction depends heavily on its implementation phase:
||AI can be implemented as an additional post-processing step to any system with access to the images (vendor agnostic)
||Only DICOM data is available. Certain information, such as image phase is removed in the earlier steps of the reconstruction and is not available for the AI algorithm.
|2- During image generation
||AI algorithms that are applied to the complex imaging data using both magnitude and phase information.
||The raw k-space data of the individual coil elements is not available for data consistency checking.
|3- During the coil element combination
||AI algorithms applied at the beginning of reconstruction chain uses all the information, also from individual coil elements. This ensures the highest data consistency and signal fidelity, enforcing that the reconstruction remains true to the acquired measurements.
SmartSpeed applies AI during the channel coil element combination phase (3), leveraging data from every individual coil element early in the reconstruction chain for maximum information content and data consistency. Via this technology, SmartSpeed offers up to 65% more resolution2 and high signal-to-noise ratio, enhancing clinician's diagnostic confidence. If desired, users can benefit from time gained with SmartSpeed by adding extra sequences for even more diagnostic information.
"SmartSpeed enables denoising the sparse under-sampling in a new way that has allowed us to increase acceleration and bring image quality to a whole new level," said Ben Kennedy, director of clinical and research MRI at Mermaid Beach Radiology in Australia. "With SmartSpeed, we've been able to use some extended applications more routinely. A good example is Philips 3D NerveVIEW. Being able to bring its scan time down to 3 or 4 minutes rather than 6 or 7, has allowed us to use it routinely and find answers that we haven't achieved before. We've introduced it in many peripheral MSK and extracranial neural applications. Now we are seeing diagnoses of neuritis in regions that we've never seen previously. It's a whole new way of looking at peripheral neurography. We've gotten an extremely positive response from our referrers."
With the reductions in scan times, Mermaid Beach has been able to:
- Fit complex cases into routine scanning times to increase diagnostic confidence
- Reduce slice thickness to increase image resolution
- Have less stress
The integration of AI-based image reconstruction with MRI technology has produced significant advancements and impressive outcomes. One of the advantages of AI reconstruction is scan time reduction. When conventional acceleration techniques are used to speed up MRI scans, there will be a compromise in image quality. This can be partly compensated by deep-learning reconstruction. However, there are limits.
One solution is to utilize an acceleration engine such as Compressed SENSE together with deep-learning reconstruction to maintain high image quality while reducing scan time. The unique balanced k-space sampling pattern used in Compressed SENSE combined with AI reconstruction at the source allows image quality to be boosted in a short amount of time.
Philips SmartSpeed is an example of how these two technologies can be used to deliver faster MRI scans without compromise. It can reduce examination time by providing MRI scans up to three times faster than parallel imaging, providing higher throughput and greater productivity for MRI departments. The time saved can be utilized to scan more patients, accommodate unplanned patients, or reduce staff overtime.
Dr. Tobias Schröter from MRT-Praxis Potsdam in Germany is already using SmartSpeed in clinical practice and mentioned that they "... achieve higher throughput and better productivity without compromising on image quality. We used to scan 32 to 35 patients per day, but now with SmartSpeed we can perform significantly more examinations in less time. We went from 160 to 170 exams per week before SmartSpeed to up to 200 per week, or about 40 patients per day."
"Since musculoskeletal (MSK) exams make up a large part of all MR procedures, these acceleration benefits have the potential to result in significant operational cost reductions," said Dr. Andra-Iza Iuga of the department of radiology at the University Hospital of Cologne in Germany.
Radiology departments can add AI reconstruction technology to their systems without bearing the high costs of replacing imaging systems. The accessibility of this technology can greatly benefit patients, ultimately leading to better health care.
Philips offers an award-winning AI-reconstruction technology not only to newly purchased MRI systems but also to most of the already installed systems. It is compatible with 97% of clinical protocols,3 so it can be used to address the imaging needs of many patients. It has been trained on an extensive set of data so that it can be used not only with 2D sequences, but also with 3D sequences, as well as for all anatomies and a wide variety of advanced contrasts such as Dixon for fat-free imaging, angiography, susceptibility-weighted imaging (SWI), and even quantitative imaging.
Furthermore, it's also compatible with non-cartesian imaging for uncooperative patients or challenging anatomies that are sensitive to motion. These technologies improve image robustness by bringing speed and image quality benefits to motion-free imaging, imaging near implants, and diffusion imaging.
"I can use SmartSpeed on patients with implants and for patients that cannot hold still, without worrying about re-scans due to motion," Schröter said. "I am astonished by the great potential of SmartSpeed. I can get very good image quality in a very short scan time."
Wide application of SmartSpeed helps MR departments to perform quantitative imaging techniques such as T1- or T2-mapping faster by decreasing acquisition time. This capability helps to bring quantitative imaging to mainstream clinical practice.
"For many years, the main limitation with most of metacontrast strategies was the acquisition durations. Now, with technologies like SmartQuant, we can accelerate fast and remove noise from the images. On a 3T Ingenia Elition system, we have optimized a 1'03 sequence with reasonable spatial resolution (1x1x5mm) and a complete brain coverage, allowing to add quantification of the brain volume. Now, this sequence can be widely used in critical care patients or in agitated degenerative brain cases,'' said Dr. Julien Savatovsky, diagnostic neuroradiologist deputy head of the imaging department at Fondation Rothschild Hospital in Paris, France.
As a result, patients can benefit from access to exciting advanced applications for brain, MSK, body, and cardiac imaging such as brain segmentation maps for assessment and triage;4 advanced visualization of cartilage structures to determination of the degradation of the cartilage; or noninvasive assessment of myocardial tissue characteristics. Reducing the exam time of quantitative imaging can help in implementing this technique more often to gain extra diagnostic information.
The comments and observations expressed are those of the author and do not necessarily reflect the opinions of AuntMinnie.com.
- Philips MI&A Insights Report MR. Q1. 2021
- Compared to Philips SENSE
- On average, measured across a sample of sites from Philips MR installed base
- Via an advanced third-party processing software: SyMRI NEURO, Synthetic MR, AB, Sweden
Copyright © 2023 AuntMinnie.com