Greg Freiherr, Industry Consultant

Greg Freiherr has reported on developments in radiology since 1983. He runs the consulting service, The Freiherr Group.

Blog | Greg Freiherr, Industry Consultant | Artificial Intelligence | February 01, 2017

The Wow Factor And The Future

3-D volume rendering

Curved multiplanar CT reconstruction and 3-D volume rendering of a heart beating 55 times per minute. Note noncalcified and calcified plaques with high-grade stenosis in the proximal right coronary artery. Effective dose was 1.8 mSv. (Image courtesy Toshiba America Medical Systems)

There was a time when cardiac cath was the only way to see what was happening in the heart. Today that can be done with mainstream computed tomography (CT) — not one souped up for cardiology, ones intended for radiological applications, ones that can freeze the motion of the human heart noninvasively. Wow.  

To appreciate where and how far we have come — and how far we might go in the future — we have to look back to where and from what we came. And pause to realize ... how ... the pace ... of progress has quickened.

Three centuries elapsed from William Harvey’s description of blood circulation in 1628 to when Werner Forssmann performed the first human heart catheterization in 1929.1 Toshiba unveiled at RSNA 2007 the first commercial CT capable of freezing heart motion with a single rotation. Less than a decade later, the company introduced a mainstream scanner capable of making stop-action heart images that are as good or better. 

We have come so far that imaging manufacturers no longer aspire to make machines capable of such “moon shot” images. They are making ones that produce previously incredible pictures efficiently, cost effectively and in ways that impose the least possible burden on patients. (Case in point: Toshiba’s latest generation of CTs produces whole-heart images at a fraction of the radiation dose required by their early predecessors.) 

And just as the future will be less than ever about spectacular images, it will be fueled by what we do not see.

 

The Binary Future

A revolution is brewing in the ones and zeros of computers — a revolution in intelligence, an artificial intelligence comprised of algorithms that distill human reasoning into machine thought.

Already these deep learning (DL) algorithms are examining medical images, determining the kind of information needed to form a diagnosis, then pulling that information together. They display images from prior exams of the patient in the same orientation, registering those images to ones in the current exam, so they can be easily compared. They extract patient demographics from electronic medical records, as well as the results of lab studies and conclusions from past reports, displaying them in the context of radiological images. 

Those algorithms aren’t in the future. They are here. Ready for market. Although they do what people had previously done, and doing so requires understanding, their tasks are routine, laborious and time-consuming. But that will change.

As the power of artificial intelligence (AI) grows, companies will reach beyond the routine. DL algorithms today are being groomed to identify trends and uncover patterns. Ultimately they may assist in actual decision-making.

Will radiologists be replaced? Absolutely not. Unless, of course, they want to be. And that is the rub.

Consider the man last summer who touted the incredible abilities of his driverless Tesla Model S. He posted YouTube videos of the car driving itself — literally driving itself. He was just a happy-go-lucky passenger, sitting in the driver’s seat but not driving — until that driverless car mistook the side of a tractor trailer for sky and crashed into it at full speed.2 

Today that man is dead, a fatality of misplaced trust. 

Stephen Hawking and Elon Musk can wring their hands all they want about the coming singularity, the moment when machines are smarter than people. And they have (“5 Very Smart People Who Think Artificial Intelligence Could Bring the Apocalypse,” Time, Dec. 2, 2014, http://time.com/3614349/artificial-intelligence-singularity-stephen-hawking-elon-musk/). They worry in the press and at conferences that machines will take over; be the end of humanity. (Never mind Musk, at the helm of Tesla, is a chief developer of AI.)

The singularity does not worry me. Much more concerning is how willing we, as individuals, are to cede control over our lives. 

I worry that, when the time comes, we will hand over our world to machines. Not reluctantly. Not after an inglorious fight. But willingly. And gladly. And we will do so with love.

 

Feeling the Love 

Americans love machines. I love machines. Why wouldn’t I? They throw snow further than I can shovel it; get me from place to place quicker than I can run; heat my home; light dark rooms; bring me amazing pictures on my television, computer and smartphone. 

Will AI be perceived as just another amazing machine built to make our lives easier?  Millions already do. They talk every day to Apple’s Siri and Microsoft’s Cortana and Amazon’s Alexa. The sale of Amazon’s Echo boomed this holiday season.3 

They know a machine is talking to them. Giving them suggestions. Helping them make decisions. Doing what they used to do for themselves. Do they care? It doesn’t seem so.

That raises the question: Where do we draw the line? 

In medicine, at least, that line is tight to the toes. At least for the time being. But soon papers will appear in respected medical journals describing the remarkable accuracy of AI; how computers make tireless assistants; how AI makes expert diagnosticians of everyone. Will someone somewhere sometime be in a rush and defer to the machine — and get away with it? Will they do it again? Will more physicians follow suit, until one day we wonder how we got where we are? 

If that happens, we will have gotten there because, after decades of trusting machines, we crossed a line. We delegated something that was uniquely human to machines ... responsibility. We got lazy ... and machines picked up the slack. Wow.

 

References

1.     Bourassa MG, “The History of Cardiac Catheterization,” Can J Cardiol. 2005 Oct; 21(12): 1011-4.

2.     “Joshua Brown, Who Died in Self-Driving Accident, Tested Limits of His Tesla,” New York Times, July 1, 2016, https://www.nytimes.com/2016/07/02/business/joshua-brown-technology-enthusiast-tested-the-limits-of-his-tesla.html?_r=0. Accessed Jan. 12, 2017.

3.     “Amazon Echo sales up 9X compared to last year, company says in holiday roundup.” GeekWire, Dec. 27,2016, http://www.geekwire.com/2016/amazon-holiday-sales-numbers/. Accessed Jan. 12, 2017.

 

Greg Freiherr has reported on developments in radiology since 1983. He runs the consulting service, The Freiherr Group. Read more of his views on his blog at www.itnonline.com. 

Related Content

CT imaging radiation dose tracking software from Imalogix displayed at RSNA 2019.

CT imaging radiation dose tracking software from Imalogix displayed at RSNA 2019. Photo by Dave Fornell.

Feature | Radiation Dose Management | March 04, 2021 | By Dave Fornell
More than a decade ago, there was an alarming, rapid rise in ionizing radiation exposure in the U.S.
GE Healthcare introduced its artificial intelligence (AI) automation features on its Voluson Swift ultrasound platform at the 2020 Radiological Society of North America (RSNA) virtual meeting. Features of this system include semi-automated contouring, auto identification of fetal anatomy and positioning on imaging. AI is seeing increasing integration in ultrasound systems from numerous vendors.

GE Healthcare introduced its artificial intelligence (AI) automation features on its Voluson Swift ultrasound platform at the 2020 Radiological Society of North America (RSNA) virtual meeting. Features of this system include semi-automated contouring, auto identification of fetal anatomy and positioning on imaging. AI is seeing increasing integration in ultrasound systems from numerous vendors.

Feature | Ultrasound Imaging | March 04, 2021 | By Dave Fornell, Editor
Recent advances in ultrasound image sy...
Philips and The Walt Disney Company EMEA announced they are teaming up to test the effects of custom-made animation, including specially-made Disney stories, within Philips Ambient Experience, a solution that integrates architecture, design and enabling technologies, such as dynamic lighting, video projections and sound, to allow patients and staff to personalize their environment to create a relaxing atmosphere

Study will explore the effects and use of animated stories and characters, including Disney favorites, as part of Philips Ambient Experience solution.

News | Magnetic Resonance Imaging (MRI) | March 03, 2021
March 3, 2021 — ...
Researchers said women who skip even one scheduled mammography screening before a breast cancer diagnosis face a significantly higher risk of dying from the cancer.

Getty Images

News | Mammography | March 03, 2021
March 3, 2021 — Attendance at regular ...
Guerbet announced the launch of OptiProtect 3S, a new range of technical services for its injection solutions. OptiProtect 3S is designed to support imaging centers in the daily use and protection of their injection solutions.
News | Contrast Media Injectors | February 25, 2021
February 25, 2021 — Guerbet announced the launch of ...
icobrain cva allows the quantitative assessment of tissue perfusion by reporting the volume of core and perfusion lesion by quantifying Tmax abnormality and CBF abnormality together with the mismatch volume and ratio
News | Artificial Intelligence | February 23, 2021
February 23, 2021 — icometrix, world leader in imaging...
Example MR images from paediatric brain tumour patients. This first column shows T1-weighted images following the injection of gadolinium contrast agent. The second column shows T2-weighted images and the final column shows apparent diffusion coefficient maps calculated from diffusion-weighted images. (a–c) are taken from a patient with a Pilocytic Astrocytoma, (d–f) are from a patient with an Ependymoma and (g–i) were acquired from a patient with a Medulloblastoma.

Example MR images from paediatric brain tumour patients. This first column shows T1-weighted images following the injection of gadolinium contrast agent. The second column shows T2-weighted images and the final column shows apparent diffusion coefficient maps calculated from diffusion-weighted images. (ac) are taken from a patient with a Pilocytic Astrocytoma, (df) are from a patient with an Ependymoma and (gi) were acquired from a patient with a Medulloblastoma. Image courtesy of Nature Research Journal

News | Pediatric Imaging | February 17, 2021
February 17, 2021 — Diffusio...