From the November 2022 issue of HealthCare Business News magazine

While artificial intelligence (AI) has been actionable for quite a while, its DL component has evolved significantly in recent years, especially in healthcare and, more specifically, in imaging. That’s because the “learning” that takes place in a DL solution is quite literal. In our own research, we’ve seen that a deep learning image reconstruction (DLIR) solution is only as good as the training it receives. Drawing on a repository of millions of low noise, filtered back projection images, these are entered as “ground truth” images into the platform. The result is imaging that achieves impressive levels of clarity.

The advantages of this approach can’t be understated. An easy, information-rich interpretation experience helps provide diagnosticians not only with the confidence they require, but also helps improve scan read times and combat radiologist fatigue. This is especially important as case totals grow, due to 6.5 million MR procedures being deferred or cancelled at the height of the COVID-19 pandemic.


We’re also seeing significant advancements in detector technology, including the delivery of two times the sensitivity in certain PET/CT platforms compared to previous versions. This can result in improvements in small lesion detectability and faster total scan times; the latter factor is especially important for those who have difficulty remaining still during imaging exams, including pediatric, neurodegenerative and geriatric patients.

Through these and numerous other technological advancements, DL is ushering in a new era of image quality. With improved visibility, pinpoint precision and faster scan times, DL is verifiably improving the scanning experience for radiologists, clinicians and patients alike.

About the author: Jan Makela is president and CEO of imaging at GE Healthcare.

Back to HCB News

November 2022 Magazine

Read more stories from the November 2022 issue

More Future of Healthcare

More Voices