Physics Will Automation Decrease Job Opportunities for Medical Physicists?

[AryaKimiaghalam]

Hi all,
Hope you are safe and doing well.

I was having a conversation with a friend of mine about the field of medical physics and its outlook. He seemed to like the field and was impressed by its career prospects. However, one worry he had was automation, in this case, it replacing the role of many physicists and reducing the number of physicists needed in each facility. The research aspect would not be automated of course but the majority of medical physicist's careers are mainly clinical. He did some research and mentioned some interesting things. Rapid development of software for initial plan verification and weekly checks, treatment planning, QA is already automated to a large extent and can be automated further; and that machine QA is being built into linac systems these days. It seems that the future career of a medical physicist is to oversee and coordinate a list of highly automated and complex processes, which could be done with much less physicists.

Nobody knows what the future entails but what could be some strong and solid arguments against an automation-related decline in job prospects of future medical physicists? (25-30 years into the future)Thanks.

 

[Choppy]

The effects of automation on job security are a concern in just about all industrial or commercial sectors, and medical physics is no exception.

I suspect we'll see the jobs of treatment planners/dosimetrists and radiation therapists change due to automation in the coming years, more so than those of medical physicists. Image segmentation and contouring, the generation of simple (and even complex) treatment plans, plan QA and delivery verification are all tasks that are being automated.

Anecdotally however, one thing I've noticed is that the medical physics workload seems to be increasing as a result. Let's say you've got 100 treatments to plan and on average they take an hour per plan - that's 100 person-hours required for the task... all performed by dosimetrists, and supported by medical physicists. Now let's say that you automate the process. You cut the person-hours involved down to 20 - that's great (if you're in management). 80% of your dosimetry workforce can now be redeployed (or what is more likely you, simply take on a greater workload). But for those 80 plans that get automated, say one or two kludge the system... the automation returns an error. It's typically the physicist who chases them, and the problems aren't always trivial. So now, now on top of all the other work medical physicists do, they have to chase the automated cases that don't work. They also have to QA the automated system on a regular basis to make sure it's performing as advertised. Their workload has increased, even if the net departmental throughput stays the same.

But of course, that's in the short-term. Eventually they'll work the bugs out, right? Data will accumulate even on the "one off" cases and the machine learning algorithms will eventually get even those correct, won't they? Maybe. But to me that seems like betting against nature to build a better fool.

If you're looking 30 years down the road, it's not inconceivable that you'll have a patient lie in a machine that will scan, plan, treat and adapt that treatment in real time. But the more complex a process gets, the more possible points of failure it's likely to have.

Plus, you still need people to hold responsibility for decisions that are being made. Something like treatment planning is a multi-objective optimization process. The computer can get you to a point where you can't really improve on one objective without sacrificing another, but at some point you have to make a decision on what you're willing to sacrifice.

There's a legal dimension to treating patients too. At the end of the day you want a person to agree with the computer that you're looking at a puppy and not a muffin, someone who can be sued when things go wrong. A person needs to hold the bag of responsibility. With medicine that's enshrined in law. Physicians have held that responsibility from the beginning, but that extends to other professions now too. Medical physicists are quickly becoming experts in automation and applications of artificial intelligence in medicine... to the point where it wouldn't surprise me if this becomes its own subfield at some point.

It's good to think about these things if you're considering a career in this direction. And what we know for sure is that the field is not going to be the same in 10, 20 or 30 years from now. I can't say with absolute 100% certainty that medical physics isn't going to disappear as a career in the coming decades. But I can say that from my vantage point, it's highly improbable.

 

[gleem]

I have seen great developments in Medical Physics since the 1970s. They reduced the workload of current activities only to create more work in the form of greater sophistication of treatments and diagnosis. Medical physicists always seemed to be trying to catch up with developing workloads. It is not likely that AI will ever be totally trusted and that human oversight will be mandated especially in human services like medicine and the judicial system.

The only things that I think will reduce the need for medical physicists are a decreasing cancer rate and a replacement of patient treatments where radiation therapy is currently preferred. Advancements in immunotherapy or robotic surgery for example. Choppy and I have had this discussion before. So far radiation therapy has been holding its own as new cancer centers continue to be built. That said it remains impossible to predict what will be the norm 20 to 30 years from now considering the rapid rate of development of technology.

 

[AryaKimiaghalam]

Thank you for your valuable insights.

 

[gwnorth]

Just an FYI but I have read several recent reports of new blood tests that have been developed for diagnosing cancer that may in the future reduce the need for ultrasounds and mammography for routine screening.

 

[gleem]

Such blood tests are aimed at diagnosing cancers earlier which will probably not hurt radiation oncology centers since earlier detection makes radiation therapy more effective.

 

[hutchphd]

The only things that I think will reduce the need for medical physicists are a decreasing cancer rate and a replacement of patient treatments where radiation therapy is currently preferred.

I think there are a host of technologies which will make R&D medical physics an exploding field. One of the last projects I worked on was an adhesive-based pap swear that required no trained professionals to either administer or screen. Such attempts to automate continue unabated. Various expert systems are already better than the best humans at diagnosis (I think breast cancer X-Ray is one such...sorry no ref) . This field is so wide-open it is difficult to know what to study. Just think about genetic markers and the information challenge there...learn lots of stuff well.

 

[gwnorth]

Such blood tests are aimed at diagnosing cancers earlier which will probably not hurt radiation oncology centers since earlier detection makes radiation therapy more effective.

The point is that the blood tests will be able to replace the need for the use of radiography for routine cancer screening. Tests like mammograms or colorectal scans for regular screening may no longer be necessary as they can be screened using blood tests. If the blood test comes back positive then the patient will be referred for additional screening and treatment if necessary, but the initial routine screening tests will not be needed.