- NIH Image Gallery/flickr
- Facebook is trying to use AI to make MRI scans up to 10x faster.
- The social network’s AI lab has teamed up with the NYU School of Medicine for a new research project.
- The company has access to 3 million magnetic resonance images from 10,000 clinical cases.
Facebook wants to revolutionise MRI (magnetic resonance imaging) scans.
The social networking giant announced on Monday a new research project that aims to use artificial intelligence (AI) to make MRI scans up to ten times faster, and has been granted access to a trove of around 3 million anonymised MRI images from 10,000 clinical cases.
In an announcement blog post, Facebook said the data it has accessed has had patient names removed, as well as “all other protected health information,” and the project is compliant with HIPAA privacy regulations.
The project is a collaboration between Facebook’s FAIR AI research lab and NYU School of Medicine’s Department of Radiology.
The efforts are a major foray by Facebook into into the medical space, as it attempts to apply its experimental AI research efforts to real-world problems. It’s an area that is rich with possibilities, but fraught with potential problems and privacy concerns. DeepMind, an AI lab owned by Google faced years of scrutiny over a data-sharing partnership with Britain’s NHS (National Health Service), and the UK’s data regulator ultimately ruled that the organisation violated British privacy laws.
MRI scans are lengthy affairs, often taking half an hour or so, while the patient lies dead still in a cramped tube. With this project, Facebook is betting that time can be cut down radically by using AI to capture less data, focusing the scanning process only on what’s important.
There are potential risks to the plan – what if in collecting less data, something is missed? – but also significant potential benefits. It means patients can take the scans more easily, and it increases the rate at which MRIs can be conducted, potentially bringing patient costs down.
“Using AI, it may be possible to capture less data and therefore scan faster, while preserving or even enhancing the rich information content of magnetic resonance images. The key is to train artificial neural networks to recognize the underlying structure of the images in order to fill in views omitted from the accelerated scan,” Facebook wrote in the blog post.
“This approach is similar to how humans process sensory information. When we experience the world, our brains often receive an incomplete picture – as in the case of obscured or dimly lit objects – that we need to turn into actionable information.”
Facebook has faced significant outrage over its privacy policies and use of customer data following the Cambridge Analytica scandal. Facebook is trying to quell those concerns with this MRI initiative, reassuring the public that “no Facebook data of any kind will be used in the project.”
There was also uproar earlier this year after CNBC reported that Facebook had been talking to hospitals about sharing data on patients, and the company subsequently said the project had been put on hold.