Practical Considerations of AI: Part 7 -- Economics

By Michael J. Cannavo, contributing writer

September 21, 2020 -- On September 3, the U.S. Centers for Medicare and Medicaid Services (CMS) approved its first reimbursement for a radiology artificial intelligence (AI) software application, awarding with a new technology add-on payment (NTAP) for use of its ContaCT stroke detection algorithm in brain CT scans. Under the NTAP, CMS will pay up to $1,040 if certain conditions are met. This is in addition to the radiologist professional fee.

Reimbursement for AI software use has been one of the biggest Achilles' heels that the industry has had to deal with, so this was huge news. But if you haven't grasped the significance of this announcement, you're not alone. Educating the AI marketplace remains another major ongoing challenge for the sector.

Ineffective market education

Michael J. Cannavo
PACS consultant Michael J. Cannavo.

A lack of financial resources isn't to blame. Flush with venture capital (VC) money, some of these companies have hired the biggest, baddest public relations and advertising agencies the VC money can buy. Sadly, many of these firms simply don't have a clue what AI is and how to promote it.

Despite many AI companies getting tens of millions of dollars in venture capital, the amount of money spent on market education by some of these firms would maybe buy lunch for the software developers who are getting top dollar to develop products no one may ever know about. Now, VC money is important for both product and company development. Few will debate that, especially in a still-developing marketplace.

But AI companies tend to invest nearly all the VC money in software development. The money is also often spent fairly autonomously, as well, with little to no oversight -- or at least not oversight by anyone who understands the market. Most VC firms would rather believe the optimistic projections that influenced them to dole out the money and not necessarily the reality of the situation in a slow-developing market.

Most VC companies also like to start to see a return on investment (ROI) at around two years. In AI, four years would be stellar and six about average -- if the company can hold on that long and the VC firms keep investing in the company. Just like what happened with PACS, excuses are inevitably given for not achieving the initially promoted goals. And these excuses are believed because the VC firms are often as clueless as many of the vendors are.

A bit harsh? Maybe, but I have seen this for several decades in the PACS marketplace with people suddenly scrambling to find new jobs when the VC firms finally pull the plug and say, "No mas." That's a shame because there were quite a few good products developed in PACS that failed due to having little to no marketing support.

No Field of Dreams

The AI marketplace isn't like the movie "Field of Dreams," in which Kevin Costner's character builds a baseball field on his farm and then several dead baseball players just show up to play.

In the AI market, nobody will show up to buy the software unless they know about the technology and why they should buy it. So market education is a must to overcome the plethora of mistakes that have already been made in the radiology AI sector.

How are these mistakes overcome? Apple's Steve Jobs said it best, "You need to start with the customer experience and work backwards to the technology. You can't start with the technology and try to figure out where you are going to sell it."

Talking about the technology and not the customer experience has been the approach AI has taken since its inception. That is very slowly starting to change, but it may be too little, too late to change some minds about AI.

Radiologists tend to view AI as a competitor, and hospitals have seen it as another cost center with value no better than the radiologist can provide. Both fail to grasp that AI and radiologists working together to help make the diagnosis are much better than either is apart.

It has always been my contention that AI is not a competitive technology to radiologists, but a complementary one. AI can help a radiologist provide a better diagnosis and offer a second opinion.

Money matters

Finally getting paid for using AI also eliminates one of the primary barriers to using the technology. Yet paranoia still seems to exist relative to AI's mere existence, let alone its use.

As an industry, AI has shot itself in the foot for way too long. So, what is the significance of the industry finally getting reimbursement for studies read by AI? It's a way for hospitals and maybe the radiologist to make money -- pure and simple. And money is the name of the game.

If a hospital can make an additional $X using a stroke algorithm -- in addition to using the radiologist's diagnostic interpretation and being paid for it -- don't you think it'll use it? Hell, yes. The same holds true for every other AI algorithm out there.

Once payment for using algorithms is approved, you are going to see this market start to take off. The RSNA is holding a contest this year to see who can develop the best algorithm for pulmonary embolism. That, no doubt, will be the next area for reimbursement, followed closely by AI for screening mammography.

A compelling argument

Whether we like it or not -- or accept it and talk at length about improved patient care and all AI's other benefits -- one of the compelling reasons to use AI involves either making or saving money. Money makes the world go round. If you don't believe me, take a look at the "most read" posts on's Forums. You'll see that many of them deal with money. Whether it is making, saving or, spending it, radiologists aren't immune to this thought process.

Don't want to use AI? Tough. You may not have a choice in the matter. Do I like the thought of radiologist assistants, physician assistants, and nurse practitioners reading studies instead of radiologists? No. That hasn't happened yet and hopefully may not at all, but realistically for some exams it may be inevitable. After all, you already have orthopedists, ob/gyns, and others conducting radiology studies in their offices, and soon emergency departments will be doing and interpreting their own ultrasound scans. How is this any different?

Now, what if you decide you just won't use the AI interpretation that is provided? That leaves the door open to the hospital setting its own ground rules, also known as, "Use it or else." And if, God forbid, you have to defend yourself in a lawsuit in which AI was right (or even wrong) and you didn't look at the AI interpretation because you didn't want to use AI? Open your checkbook and update your CV ... if you can find malpractice insurance that is.

Not about you

I know what I say isn't popular. Rick Warren started his book "The Purpose Driven Life" with four words that define where we are in medicine today as well -- "It's not about you." I don't like it any more than you do, but it is what it is. We are no longer in control of our own destiny. As technology advances, we become less and less in charge of it.

We need to learn more and understand and accept how AI works with an enterprise imaging system/PACS and what the future holds for both technologies. PACS is here to stay, and so too is AI. Some dismiss the financial aspect of AI, but making up to $1,040 per case by using an algorithm is far from a "zero-sum gain" and a "waste of money," as some have said.

Once again, it shows that those who fear AI just don't understand how the technology and application of it therein really works. Nowhere does it also say a radiologist will get paid less for doing the interpretation either -- at least not as of now. It just says that there will be additional payment for the use of AI.

Future financial impact

However, as has been seen in the past, once the radiologist's involvement in the diagnostic interpretation process is determined relative to that of the AI algorithm, then I'm sure the relative value units will be adjusted accordingly. If the AI companies are smart and position AI as an adjunct to the diagnostic interpretation process -- and not as a replacement -- they may be able to minimize any financial impact it will have on the radiologists. Will that happen? That again is anyone's guess.

Radiology tends to focus on the radiology world and not beyond it. Hospitals are also looking closely at how AI can assist cardiology, pathology, and even the emergency department physicians as well. Stroke is the No. 1 cause of adult disability in the United States and the No. 5 cause of death. Research has found that every minute a stroke patient's treatment is delayed leads to an additional one week of disability. This is why suspected strokes are stat cases requiring on average no more than a 15-minute interpretation. Although the AI algorithms being offered today are far from perfect, an average 90% accuracy rate beats 0% every day, especially in the emergency department, where seconds count.

Rick Warren also made another interesting comment in his book: "We are products of our past, but we don't have to be prisoners of it." If we are willing to accept the change and embrace AI just as we accepted that same change with PACS, then there remains hope. Without it, well...

Michael J. Cannavo is known industry-wide as the PACSman. After several decades as an independent PACS consultant, he worked as both a strategic accounts manager and solutions architect with two major PACS vendors. He has now made it back safely from the dark side and is sharing his observations.

His healthcare consulting services for end users include PACS optimization services, system upgrade and proposal reviews, contract reviews, and other areas. The PACSman is also working with imaging and IT vendors developing market-focused messaging as well as sales training programs. He can be reached at or by phone at 407-359-0191.

The comments and observations expressed are those of the author and do not necessarily reflect the opinions of

20 ways to prepare for the 'new normal' in radiology
We are living in what will soon be known as the "new normal." With this new normal comes changes that we have to either accept and adapt to or face the...
How can digital imaging help in the coronavirus crisis?
Could digital technologies like PACS and enterprise imaging help healthcare professionals manage viral outbreaks like COVID-19? Industry consultant Michael...
The 2019 PACSman Awards: All about AI
CHICAGO - RSNA 2019 was all about artificial intelligence (AI) -- now if only we could make heads or tails of it, writes Michael J. Cannavo, aka the...
Video from RSNA 2019: The PACSman takes on artificial intelligence
CHICAGO - Long known for his iconoclastic views on digital image management, the PACSman -- Michael J. Cannavo -- also has a few thoughts on the looming...
Practical Considerations of AI: Part 6 -- Ready, fire, aim
In part 6 of his Practical Considerations of AI series, PACSman Michael J. Cannavo shares his thoughts on why the market for radiology artificial intelligence...

Copyright © 2020

Last Updated mf 9/18/2020 4:18:35 PM

4 comments so far ...
9/21/2020 10:40:54 AM
           I got an e-mail from one of the AI vendors asking me “Isn’t it unprecedented to reimburse for a specific vendor’s algorithm?” Unprecedented? The amount of phone calls and emails I have gotten on Part 7 certainly has been unprecedented for sure. And since no vendor has ever gotten reimbursement before, to that end reimbursement for a specific vendor’s algorithm can be considered unprecedented. I think the bigger question being asked is why only one vendor’s algorithm and not all the 510-k approved stroke protocol algorithms that available out there? I’m not asking why just happy even one got approved.  .
          When my kids were young they would ask me why my oldest son got something the youngest didn’t (or vice versa).  I would play with their heads a bit and tell them its because "I like your brother better, that’s why"…but then I realized if one of them grew up a serial killer because of own stupdiity that would not be cool so I stopped doing that after just a few times. I can’t answer why one vendor is approved and not another but I’m sure the answer is much better than “I like them better...”.  
         Keep in mind this is an add on payment that only deals with inpatients on Medicare so it the end add on payment is limited in who gets it. It's also not a CPT code (that remains a ways out yet) but at least it's s start ….and the beleaguered and developing Ai industry will take whatever it can get.
          Want to learn more? Hear what AI vendors are doing right, doing wrong, market dynamics and more in “The Challenge of Integrating Enterprise Imaging with AI” webinar this Weds 9/23 at 10 a.m. The event is FREE. To register:
Mike Cannavo
(407) 359-0191 (Office)
(407) 247-7345 (cell)

9/21/2020 11:23:05 AM
Mike, you said:
In the AI market, nobody will show up to buy the software unless they know about the technology and why they should buy it. So market education is a must to overcome the plethora of mistakes that have already been made in the radiology AI sector.
Are we sure about that? Look at the hundreds of millions of bucks thrown at Nanox, which appears to be less than kosher so far...

9/21/2020 12:15:53 PM
Ah yes Nanox....Thernos 2.0....
There have been hundreds of millions of dollars thrown at AI as well- but dispersed among numerous companies...but at least we are seeing SOME products developed even though no one may know about it yet....Maybe that should have read " NO REASONABLE (OR EDUCATED) PERSON" INSTEAD. 
Don't even get me started on VC's. The editors at AM kept me from getting shot by editing out some choice observations I made about the VC's. Let it suffice to say they really do need some fresh eyes looking at the investments they make both before and after with someone constantly providing input as to how the VC money is being spent... not blind trust or listening just to the advisory board. 
I love the comment made by one VC "about  how exciting it is supporting a company that improves patient care and saves lives on a daily basis. Pulease.....Ever seen Shark Tank? Jerry McGuire? Yes, it's great that we can help patients  but most VC's I know resonate Rod Tidwell having his "chat" with Jerry McGuire.
      "I wanna make sure you're ready, brother. Here it is: Show me the money. Oh-ho-ho! SHOW! ME! THE! MONEY! A-ha-ha! Jerry, doesn't it make you feel good just to say that! Say it with me one time, Jerry."
Show me the money....
In light on the Nanox's of the world, AI needs to be like Hebrew National- they, too, need to answer to a higher authority...


9/21/2020 5:07:00 PM
          An astute radiologist who heavily involved in the AI industry and is up to date on the CMS action relating to payments for using the viz-ai’s stroke algorithm, protocol provided me with some excellent information that bears sharing with AM readers.:
1)   The Viz.AI payment announcement technically is not a “'real reimbursement” but an add on payment. I clarified this point already but it bears repeating. Whatever you call it the NTAP program is still a positive step in the right direction,
2)   CMS has said they will provide reimbursement ONLY for the following narrow group of patients:
                   1. Medicare patients  (approx 50-70% of stroke work ups)
                   2. Code stroke patients that had a CTA (angiogram) and
                       the  software was used to determine if a stroke
                       occurred or not. No clarification was made if a CTA
                      was done and a  stroke not found. I assume you would
                      get the add on payment either way.) 
                   3. Patients  that showed a financial loss compared to the
                       DRG that was billed (approx 50%).will get this
                       payment. Specifically, 65% of any loss will get
                       reimbursed up to $1040 per patient. What this basically
                       says is CMS will only provide a payment for patients
                       where there was a financial loss and only up to a max
                       amount. In fact, the  program is more about CMS re-
                       adjusting their DRG  payments year-to-year. They will
                       then look at this payment each year for 3 years (and
                       each year it can be changed from $1040). After that it
                       is anyone' guess as to what will happen to this
                       payment. The idea is that , based on the data, CMS
                      would just adjust the stroke DRG payments and not
                      really 'reimburse' the AI since it would be incorporated
                      in the  DRG. 
          No one knows what the outcome of the program will be. This is not a reimbursement like many are used to (i.e..for CAD or other imaging studies). Also ,it depends on how much Viz charges for their subscription model, which may  eat up any real savings here. That said,  any payment is  better than no payment, but we have a ways to go...
          I am truly indebted to this physician for sharing this information.  I just reinforces one of the key points I made in the article-  it’s all about market education. Finally showing money coming in for using AI was promoted as a “ground breaking ruling” and the word “reimbursement” used more than once in print, including by myself.  In fact, were it not for a webinar that the individual who shared this data with me  attended we would know nothing rldr. I like to think I keep up to speed on what is going on in the industry but  knew nothing of  a webinar or would have sat in on it..I’m sure many others in the industry and those who care to look more closely at using this echnology are all in the same boat,.
          It all comes back to two words- market education. That is and always will be what it’s all about…
The newly educated PACSMan