How AI will make help in making MRI scans faster

By arvind |Email | Aug 22, 2018 | 6888 Views

When was the last time you were excited for an MRI scan? It's a long, time consuming and a discomforting process which gets worse if your health is worse. While MRI scans can last 15 minutes, an MRI scan of the heart can take an hour or more. At present the process is slow machines are slow to create a usable scan or image, a radiologist needs to have a clear image to interpret. To create a clear image the machine has the capture all the data necessary which takes time.
Facebook and NYU School of Medicine have collaborated on this work and have announced fastMRI, the process will use AI to make MRI scans almost 10 times faster. Improving the process of MRI will make the process faster and making the technology easily available to more people. 

Daniels Sodickson, a professor in the department of radiology at NYU School of Medicine said," using AI, it maybe possible to capture less data and therefore scan faster while preserving or even enhancing the rich information content of magnetic resonance images."

How will it be different from present MRI's?

The focus in the beginning is to change how these MRI machines work. At present, the scanners collect all the raw numeric data and then turn the data into an internal body structure image. That is reason behind the long duration of scan, while some scans have less data to collect others had large data to collect to create an image. The main focus is to train the neural network with the use of AI so that it could identify the structure of the image that is lurking so a complete image can be created; views can be left when doing this accelerated scanning. NYU teams working on this project have been able to generate high quality image from far less data using the artificial neural networks. 

How it all happened?

CAI2R have been working on the concept of using AI to achieve faster MRI scans since 2016. There studies in the past have suggested that the scan times of these scans can be reduced fairly more. But what they were lacking was additional AI knowledge and large scale computing resources that when FAIR (facebook Artificial Intelligence Research) who were looking for projects that can use AI to have their use in the real world. That's when FAIR can into the image, for them CAI2R image reconstruction fitted well with FAIR's deep learning expertise and they say it as an opportunity to combine both.

What is in the future?

The partnership aims to create high value images that provide medical visualization with new capabilities that they can use to help humans with health benefits.  Right now the project is working on the MRI technology. In the future it is expected to open its arms to other medical applications. FAIR wrote "in the interest of advancing the state of art in medical in medical imaging as quickly as possible we plan to pen sources this work to allow the wider research community to build on our developments." They further wrote "as the project progresses, facebook will share the AI models baselines and evaluation metrics associated with the research and the NYU School of medicine will open sources the imaging data set. This will help ensure the work's reproducibility and accelerate adoption of resulting methods in clinical practice."
facebook code

Source: HOB