Title and Logo

First published as:
Maverinck –
What will AI really mean for radiology?
21 November 2018.
Aunt Minnie Europe


This version has been extended with some additional references.
The version published in Aunt Minnie Europe was shortened and edited to make it fit.


Rinckside
ISSN 2364-3889

Rinck PA.
Will artificial intelligence increase costs of medical imaging?
Rinckside 2018; 29,6: 15-16.
Read the Print Edition (PDF)



Will artificial intelligence increase costs of medical imaging?

t present, many attendees of conferences and courses see medicine and radi­ology through the lens of fanatic devotion to artificial intelligence. Its promoters promise earlier and more reliable diagnoses, fewer examinations and procedures to establish these diagnoses, less morbidity, perhaps even mortality, altogether a better outcome for the patients, and lower costs for the health care system.

Since it's the talk of the town, of scientific journals, newspapers, the internet, conference sessions, it must be good, or so the theory goes. The neighbor has it, let's get it too. And one can even play with it: men against machine, sometimes even on stage in front of a congress audience.

Nobody mentions that AI is seductive but an unsettled and immature technol­ogy that requires permanent updates.1 It will be a cash cow for developers and the industry, and add to the exploding costs of medical imaging: Nothing is stable, nothing is really reliable; there is permanent change that foils the stability and validity of radiological diagnostics.

AI software will never be a final product. The programs always need to be up­graded and updated. Much in radiology that is subsumed in artificial intelligence today had an earlier life going by another name.

Some years ago I cited Dr. Donald A. Berry from the M.D. Anderson Cancer Center in Houston, who summarized his experience with artificial intelligence in mammography. At that time AI was still called “CAD” in radiology, “computer-as­sisted diagnosis” (or “detection”) [1].

"An argument for the use of CAD with film or digital mammograms is that it will get better over time. Fine. Researchers and device companies should work to make the software ever better. But this should happen in an experimental set­ting and not while exposing millions of women to a technology that may be more harmful than it is beneficial [2]."

Pressures on radiologists

AI software is like coffee in cap­sules – expensive, but you never know what's really inside. However, there are social, economical, and political pressures to conform and to purchase certain digital devices. Radiologists are also considered consumers and told by others what is good for them.

Salespeople are already salivating over AI sales contracts. Many business models rely on artificial intelligence to facilitate tasks so much that we no longer want or can do without it.

ruler black

Salespeople are already salivating over AI sales contracts.

ruler black

Health care, the commercial part of medicine, avoids any accountability; new techniques and methods are introduced at random, praised to the skies enthusiastically, partly militantly by technocrats and paid experts. The shelf life of many of the new fashions and products is approximately two years. That's at least the interval we were told by two representatives of CO­CIR, the European Trade Association representing the medical imaging indus­tries, invited to a meeting on the future of radiology at WHO in Geneva some time ago.

Therefore, they said, outcome studies are irrelevant – all digital pro­cedures or equipment you buy today will be hopelessly outdated in five years, after limited use. Still, they claim that these very same products will increase productivity, one of the central themes favored by commercial salespeople and hospital managers.

At the time of the introduction of x-ray CT, 45 years ago, the overall costs of medical imaging were between 1% and 3% of overall health care expenses; to­day just the yearly worldwide sales of diagnostic and therapeutic imaging equip­ment amounts to 100 billion euro; the sales increase by 5% annually [3].

Artificial intelligence is a mix of the virtual digital world and the real world. Intel­lectually, it is a step backwards. It's a shift of knowledge and assessment of col­lected and processed data from the human brain to a black box digital system – and the blind reliance on the correctness of this system. But it doesn't provide a rational, independent opinion. It also creates an addictive dependence because people will tend to totally rely on it. Already voices are raised claiming that the use of AI will lead to de-skilling of the workforce. Immature and costly technologies shouldn’t be used in medicine.

A recent Italian paper stresses that the processes of medical device decision-making are largely unpredictable and points out that there are major differences between Europa and the United States:

"Generally, while in the U.S. AI the technology sector prospered in a permissionless innovation policy environment, in the EU decision-makers adopted a different policy for this revolutionary technological branch. Certainly, swifter approval of AI medical devices helps generate revenue for manufacturers, and physicians may benefit from having more tools at their disposal. But the final goal of bringing new devices to market should be to improve prevention, diagnosis, treatment, prognosis of diseases with a potential positive impact on patient outcome. Therefore, systems for approving new medical devices must provide pathways to market for important innovations while also ensuring that patients are adequately protected. To achieve these goals, the EU and the U.S. use different approaches [4]."

The Canadian perspective

Arguably the best and most balanced review paper on AI in radiology was the white paper published by the Canadian Association of Radiologists in May 2018, considering the pros and cons of the introduction of AI in diagnostic imaging. It is worthwhile reading [5]. The authors are realistic and down-to-earth in terms of applications and development:

Practicing radiologists need to understand both the value, and the pitfalls, weaknesses, and potential errors that may occur when an AI product performs image analysis. While these algorithms are powerful, they are often brittle, and may give inappropriate answers when presented with images outside of their knowledge set ... an algorithm-evaluating brain CTs may work perfectly for long stretches, but then a new software upgrade to the CT occurs, or a new CT machine comes on-line, and all of a sudden, the algorithm produces faulty results.

AI is a mix of the virtual digital world and the real world. It's a shift of knowledge and assessment of collected and processed data from the human brain to a black box digital system -- and the blind reliance on the correctness of this system. But it doesn't provide a rational, independent opinion. It also creates an addictive dependence because people will tend to totally rely on it. Already voices are raised claiming that the use of AI will lead to deskilling of the workforce.

As the Canadians remark: "Currently, there is no evidence in the literature that AI can replace radiologists in day-to-day clinical practice. However, there is evidence that AI can improve the performance of clinicians and that both clinicians and AI working together are better than either alone."


spaceholder blue

References

1. Rinck PA. CAD as CAD can. Rinckside. 2011; 22(9): 17-18.
2. Berry DA. Computer-assisted detection and screening mammography: Where's the beef? J Natl Cancer Inst. 2011; 103 (15): 1139-1141.
3. European Coordination Committee of the Radiological, Electromedical and Healthcare IT Industry (COCIR). Our industry. Accessed 20 November 2018.
4. Pesapane F, Volonté C, Codari M, Sardanelli F. Artificial intelligence as a medical device in radiology: ethical and regulatory issues in Europe and the United States. Insights Imaging. 2018; 9 (5): 745-753.
5. Tang A, Tam R, Cadrin-Chênevert A, et al. Canadian Association of Radiologists white paper on artificial intelligence in radiology. Can Assoc Radiol J. 2018; 69 (2): 120-135.

spaceholder blue003 spaceholder blue003

VirtualCampus

00-f1

00-f2

00-f3

00-f4

00-f1

00-f2

00-f3

00-f4

00-f1

00-f2

00-f3

00-f4

00-f1