Benoit Doche de Laquintane, MD

Dr. Doche de Laquintane uses Canon's Vantage Galan 3T system with eight coils, 32 elements, gradient technologies, and a slew rate of 200. He also uses Advanced intelligent Clear-IQ Engine (AiCE) Deep Learning Reconstruction software and other advanced acceleration techniques, like Compressed SPEEDER and MultiBand SPEEDER.

AiCE is the world’s first Deep Learning Reconstruction technology available MRI for scanners. It's very easy to use and seamlessly integrates with our routine workflow.

When you compare a conventional filter with AiCE, you can clearly see the difference. There are anatomical details in the conventional filter’s subtraction, so you lose some clinical information during denoising. However with AiCE, there is only noise in the subtraction. There is no relevant clinical information. So the signal to -to-noise ratio is better.

Here’s an example of MultiBand SPEEDER acceleration on DTI. It took two minutes 22 seconds without AiCE and one minute 16 seconds with AiCE. Nearly 50% of the time was saved. In efficient situations , AiCE saves between 27 and 55%.

A 3D FLAIR that took four minutes 40 seconds before now takes two minutes 15 seconds. A susceptibility FSBB that took three minutes 38 seconds before now takes one minute 41 seconds. And with 3D T1 MPRAGE, what took two minutes 42 seconds now takes one minute 42 seconds. So this standard protocol for stroke takes less than seven minutes.

Setting the denoising system to 1 is very conservative. At 1.8, you start pushing the system toward its limits. When you start using AiCE, you have to try settings for yourself and decide on the acceptable balance of quality and time savings.

Another acceleration technique is Compressed SPEEDER, which undersamples the K space . You use a regularization factor and the point is to separate the redundant information, or tissue signal, from the auditory information, or noise. It's a trade-off because if you push the regulations factor too high , you’ll have less tissue signal and image quality will decrease.

Just to show you how good the resolution is, here’s an 84-yearold woman with right arm hemiparesia and right facial weakness. You can see there is atrophy in the palatal lobe and a small recent subcortical infarction in the superficial territory of the middle cerebral artery.

The last example is a woman with an episode of acute dizziness three months ago. You can clearly see a very small left lacunar cerebral infarct. It strikes a good balance of time savings and resolution.

You will improve your patient workflow as well as your diagnostic quality and have more time available for more sequences or patients. Patients will spend less time waiting.

You can more easily do advanced applications like spectroscopy, perfusion T1, perfusion T2 and ASL by using these various acceleration techniques. You can also combine DLR with CS and probably use synthetic MRI for 3D in the future. The idea of all this is to achieve 7T quality with a 3T magnet.

You can improve the workflow at 1.5T and 3T. For example, on 1.5, we tried to do four examinations per hour by alternating two examinations with contrast and two without contrast. And on 3T, we tried to do an advanced application in 30 minutes followed by two quick examinations in just 15 minutes. As you can see, the routine protocol is around six minutes 49 seconds. For epilepsy, it's around eight minutes 37 seconds.

So the takeaway message is that a faster workflow and improved diagnostic confidence in stroke follow-up is now possible with new acceleration techniques due to Deep Learning Reconstruction, Compressed SPEEDER, DLR AiCE, and CS at 2D, as well as 3D soon, which we hope will become faster in the future. //

See more topics in this webinar.