The Johns Hopkins University in Baltimore, Maryland, USA, is one of the world’s leading centres for clinical research. It has collaborated with Canon Medical Systems (hereinafter "Canon Medical") for many years. The partnership has been strengthened recently, particularly during the COVID-19 pandemic with a team led by Dr. Joao Lima, Director of Cardiovascular Imaging and Professor of Medicine along with Dr. Chia Liu, Senior Clinical Scientist at Canon Medical Systems Corporation, and a number of other international experts to investigate the wider clinical effects of the disease in a series of MRI studies. Their ground-breaking work has been made possible through the Vantage Galan 3T MRI System along with its artificial intelligence (AI)-supported technologies, including Precise IQ Engine (PIQE), Canon Medical’s most advanced Deep Learning Reconstruction technology.

Canon Medical has a long history of collaboration with Dr. Lima. It began in 2002 at an American Heart Association meeting, when Dr. Lima and his team saw images of a CT-guided arterial stent being placed. The team was so impressed that they began to invest in various Canon Medical equipment.
A few years ago, Canon Medical decided to take space in the Johns Hopkins biotech park and installed a Vantage Galan 3T MRI System.

“The Unit was created to pursue the goal of a totally contrast-less MRI. That was the original idea, and we are still pursuing that,” said Dr. Lima, “However, our team and Canon Medical’s technology have produced work that has made an impact in the field. We have Canon CT systems, as well as MRI and other equipment.”
When the COVID-19 pandemic emerged, Dr. Lima’s team wanted to gather further insight into what happens after the acute phase of the disease, and they began a study into the morphological and quantitative analysis of COVID-19 sequalae.
“We developed a protocol in search of sequalae in patients convalescing from COVID-19, which includes the brain, the liver, the heart, the lungs. With COVID-19, it’s most important to keep the patient in the scan for the shortest possible time, so the protocol includes T1, T2 mapping on the heart, T1 mapping on the liver, with T2* (ultra-short TE) imaging on the lung,” explained Dr. Lima. “Such a complicated protocol can now be done, because of big developments in Artificial Intelligence (AI). And I think the impact of AI in MRI might be even greater than in CT, because in MRI, you are at the centre of a very tough triangle. You have to compromise between scan time, signal-to-noise ratio (SNR), and spatial resolution, and your protocol is always tailored to trade-off between these components.”
In cardiac, a lower acquisition matrix and a higher acceleration factor are traditionally used to deal with the ‘triangle’.

Dr. Lima believes that PIQE could take protocols to a whole new level.
“We are redesigning what an MRI protocol should be, based on PIQE,” he remarked.
“There are two very important components - the deep learning component, which takes an image in through a convoluted neural network, which is trained with very high-resolution images, which is particularly useful to denoise images. And it works very well.” He continued. “At one point, we only had that piece of the tool, but now, we have the whole tool, which also comprises zero filling, which is a technique that we know in MRI to improve the spatial resolution. And then, we send the output back to the deep learning algorithm to clean for the artefacts that are generated with zero filling.”
In cardiac protocols, higher accelerations can be applied by using PIQE and breath hold can be reduced.
“We're really tied to the breath hold in cardiac imaging because motion is a big issue. Many of our patients can't hold their breath for very long. Therefore, that becomes a limiting factor,“ said Dr. Lima. “If you can reduce the breath hold from nine to six seconds, it is a big advancement. And we can do the things that we generally do but do them much easier. And the result is that it's consistent.”

“Now we are using PIQE not so much to enhance throughput but to convert patients that we couldn't easily image before to obtain images that we can read easily and that are consistent,” he added.
“For T1 mapping, the effect of PIQE is basically to allow regional measurements because it creates homogeneity in the height. It doesn't change as much the accuracy of the method, but it does enable you to make measurements at different spots in the cardiac tissue,” said Dr. Lima. “On the other hand, for T2 mapping the difference is striking. You can see that because T2 imaging generally means imaging with a very low SNR because T2 imaging provides by its nature a smaller signal. That is where the opportunity for AI is, because if you can transform imaging with low signal into very clear images, that's a huge achievement. We get the same thing with Bright Blood which is a T2 based image. It is much better with PIQE. We even use it also outside the heart in the vascular system to measure cardiac output, to measure pulse wave velocity in the aorta.”

What Dr. Lima appreciates in many applications is that PIQE adds reliability.
“The end quantification is very reproducible because the algorithm really reduces noise and increases resolution,” concluded Dr. Lima. “It's a very clever algorithm.”