Technology companies are exploring new avenues for their expertise in artificial intelligence (AI) can be put to better use. The world’s largest social networking company, Facebook Inc. (FB), has announced working on a research project related to medical imaging. The initiative is being jointly launched with a team of doctors in the radiology department of New York University's School of Medicine.
At present, a magnetic resonance imaging (MRI) scan takes anywhere between 15 minutes to an hour. MRI is a type of scan that uses strong magnetic fields and radio waves to produce detailed images inside the body and is a popular method of diagnosis. The joint project aims to make MRI scan times up to 10 times faster. If successful, it will become a game changer, especially in cases where quick diagnosis and turnaround is required for timely treatment. Additionally, the shorter time cycle will also free up the MRI apparatus to be used by many more patients; currently, many MRI facilities have waiting lists of days or weeks.
Facebook engineers belonging to the Artificial Intelligence Research (FAIR) group plan to use neural networks for the innovative project called as fastMRI. Neural networks is a series of algorithms that seek to identify relationships in a data set via a process that mirrors the working of a human brain. The researchers will utilize around 3 million MRI images of the brain, liver and knees sourced from 10,000 different medical cases as available with the NYU School of Medicine. To ensure data security and necessary anonymity, all details of the involved patients are removed from the medical images. No data from Facebook social media profiles are being used.
Attempts to Speed Up MRI Scans
The team will first study how MRI scan is performed in the current process, where various body scans are combined to make suitable images. The next phase involves assessing if AI can deliver similar or better results more rapidly with smarter scans that capture and process less data. “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,” researchers involved with the project state. Initial findings have revealed positive signs: AI was successful in generating suitable scans from less data.
The Menlo Park, California-based company has been making strides in the field of AI and has expertise in data as well as image processing. It has used AI to contain the spread of illicit content to a significant extent on its network, something that would have been hard to achieve with human operators and standard programming. (See also: Facebook Using AI to Combat Terrorist Propaganda.)
Last year, the company closed down a project that was attempting to train automated bots to negotiate, though it has used AI for sucessfully rendering translations on its platform. (See also: Facebook's AI Translates 9X Faster Than Rivals.)