Philips uses AI to improve MR imaging

  • July 19, 2022
  • Steve Rogerson

Philips’ SmartSpeed artificial intelligence (AI) powered magnetic resonance (MR) acceleration software has received US FDA 510(k) clearance.

Adding AI data collection algorithms to Philips’ existing SmartSpeed compressed sense MR acceleration engine delivers higher image resolution with a claimed three times faster scan times and virtually no loss in image quality, representing a step forward in diagnostic confidence and MR department productivity.

With personalised treatment for complex diseases such as cancer increasing the need for high-confidence precision diagnoses, coupled with soaring caseloads due to aging populations and high levels of clinician burnout, radiology departments are under increasing pressure to improve performance, productivity and profitability.

“Philips’ AI-based SmartSpeed reconstruction is the new benchmark among acceleration techniques for us,” said Grischa Bratke, radiologist and expert in musculoskeletal imaging at the University Hospital of Cologne in Germany. “It improves on the company’s existing compressed sense in all aspects and allows a reduction in scan times with excellent image quality and diagnostic confidence.”

SmartSpeed is the latest addition to Philips’ expanding portfolio of AI-driven, smart, connected imaging and smart workflow products. It helps achieve the quadruple aim of improving diagnostic outcomes, enhancing patient and staff experiences, and reducing healthcare costs overall.

“By taking the quadruple aim as a starting point to develop and deploy AI, we have a tremendous opportunity to address the inefficiencies and growing demand that our customers and partners face every day,” said Kees Wesdorp, chief business leader of precision diagnosis at Philips. “Powered by AI, our leading technologies and smart connected imaging help turn data into actionable insights to increase diagnostic confidence and help improve clinical outcomes for patients.”

The AI-powered MR software increases resolution up to 65% and achieves up to three times faster scanning times compared with conventional MR scans. Coupled with its support for 97% of current clinical protocols – including contrasts and diffusion weighted imaging, and even quantitative imaging such at T1 or T2 mapping for brain, liver, heart or musculoskeletal (MSK) imaging –SmartSpeed helps increase diagnostic confidence in a wide range of patients, including the most complex cases.

The deep learning technology is applied at the source of the MR signal and seamlessly integrated with Philips’ compressed sense speed engine to deliver the higher resolution and high signal-to-noise ratio needed for detailed images, small lesion detection and enhanced diagnosis.

An AI reconstruction algorithm at the front-end of the MR signal removes noise and preserves details while enabling a k-space data consistency check for trustworthy AI.

The compressed sense adaptive AI technology was co-developed by Philips and several of its academic partners. It won the 2019 Fast MRI Challenge hosted by Facebook AI research and New York Langone Health.