How deep learning could change diagnostic imaging forever


Computational imaging is a passion for Fabein Beckers. As the CEO of Arterys, Beckers thinks he can revolutionize how doctors on a global scale work with imaging diagnostics.


The Arterys system was launched as part of GE’s ViosWorks.

In that vein, Beckers and his team launched Arterys, which enables clinical visualization and accurate quantification of blood flow inside the body. It measures 7 dimensions of data (3 in space, 1 in time, 3 in velocity direction) to improve diagnoses in cardiac medicine. The technology relies on cloud computing using artificial intelligence and deep learning to enhance the abilities of physicians in a safe and expansive way. The Saas system is poised to revolutionize the field of computational imaging using the same technology that has transformed Google and Facebook. Today, these sites feature face recognition, highly targeted advertising  and live data transfer – all effects of deep learning.

Paired with optical intelligence technology, said Beckers, there is a quantum leap in applying these tools to medical imaging.

How Arterys solves a problem

The technology is not an attempt to replace technicians or physicians, noted Beckers. The goal is to assist the physician processing and dealing with the exam of a heart during an MRI. In the early days of imaging, a radiologist might produce one X-ray. Now an average medical image of a patient produces at least 2,000 images.

In a typical MRI, images are captured in several slices from the top to the bottom of the anatomy. During the process, doctors stand there and try to understand the volume of the ventricles when the heart is open and the capability of the muscle to pump blood. The physician finds, tracks, and contours objects, in the image and takes measurement after measurement. It is a very tedious process that can take up to an hour and a half per patient.

“It is definitely not a good use of a physician who has a decade of training and who already has so much pressure on cost reduction, on-time, and on volume,” Beckers said. “I would rather have them focus on the clinical question, not draw circles on software all day long.”

The current process is also subject to inconsistency across caregivers, which means obtaining high-quality data is difficult. Right now, Beckers said, the field is so fragmented that gathering meaningful data across players is hard.

In the Arterys system, when a heart is opened by the physician everything that would be manual, e.g., the contouring of the ventricle, becomes automatic and therefore it is extremely accurate. There is still an override mechanism.

“Our mission is truly to augment physicians,” Beckers said. “If they feel that the tool our software has created is not perfect and there is something that needs to be changed, we can always avert and modify it. By doing that, we are creating something that is better every single time.”

Arteys captures those data changed by the physician and applies it to make the AI software smarter and work better for individual physicians.

Toward predictive health

The technology is particularly suited to create predictive models, Beckers said. “Imaging is a great path to enable a new way of healthcare being predictive, quantified and consistent. Because there are only 3 vendors, the data is highly standardized,” Beckers noted. He said that the Digital Imaging and Communications in Medicine (DICOM) standards have enabled structured data and the establishment of interoperability product codes. All the medical imaging industry needs it a pathway for bringing AI into play.

The cloud system enables predictive analysis from one patient to another and allows collaborative work from physician to physician. A system that promotes collective aggregation builds a better system for everyone. “The AI gets smarter, the doctors get better results, and they all collaborate to really create a level playing field,” Beckers said.

Beckers noted that HIPAA concerns are alleviated because Arterys has included technology that strips out any personal health data from the scan and only uploads the image to the cloud. “The cloud will never host any PHI,” Beckers said.

Now and later

In December 2015, Arterys was launched as part of GE Healthcare’s ViosWorks, which aims for comprehensive cardiac MR diagnostics. Beckers absolutely sees the technology as a platform for many different diseases, particularly for cancer.

“We have a joke that my Uber driver has more access to predictive technology than my physician,” Beckers said. “The medical field and technology have been separated, but they are getting closer together.”

And for Beckers, closing the Venn diagram of health technology and software technology could not come any sooner. “Everything in medical is so managed, so tedious, so subjective. We are trying to help physicians gain access to data-driven medicine.”

Such data available enables informed decisions that are best for the patient, and it alleviates manual tasks that don’t leverage physician’s expertise, training and skill set.

[Want to stay more on top of MDO content? Subscribe to our weekly e-newsletter.]

Want to stay on top of MDO content? Subscribe to our e-newsletter.

Speak Your Mind