AI should benefit "Little RadOnc"

[CurbYourExpectations]

---

Members do not see this ad.

Artificial intelligence will continue to advance at a pace that exceeds human capability. This trajectory is inevitable and unequivocal. Benefits to patients lie within harnessing AI within “Little RadOnc”. Community or smaller-scale radiation oncology practices. The most significant AI breakthroughs will likely originate from large institutions and research groups, their downstream optimization and practical application over the coming decades will increasingly empower smaller practices through software related advances.

It is a given that software advances will continue to outpace hardware advances. Creators of hardware will be opposed to this. A cautionary parallel can be drawn from the gaming industry leaders that create new systems with exclusive games/software, even with old hardware that could have ran a similar version. If this happens in medicine it could exacerbate disparities between small radonc and big radonc. We need a growing level of scrutiny applied to hardware changes, especially as they relate to data integrity, algorithmic fairness, and other nuances.

I think AI advancements could empower smaller, resource-limited practices to demonstrate equality in care and decrease need of sending patients to larger centers. Acute attention and leverage of validated AI tools may provide comparable outcomes without necessitating referrals to high-cost tertiary centers. The benefit of this? Patients remain closer to home, we don't waste money on machines that don't have true proven benefit based on biased study parameters or limitations of making software applicable to older machines in orders to build profit on hardware.

Embrace and discuss advances in medicine. Scrutinize new expensive machines that can increase billing. Prevent a need for referring by providing similar care, this will be made continually easier with software, in spite of the expenses of what could be unnecessary hardware.

 

[TheWallnerus]

Your love of the halflings' leaf has clearly slowed your mind 😉

 

[CurbYourExpectations]

Your love of the halflings' leaf has clearly slowed your mind 😉

You're someone I would hope would speak and give input.

 

[Palex80]

I asked a resident of mine how he would treat a challenging case, his answer was spot on. I congratulated him. He said, he asked AI for the answer.
🤣

 

[CurbYourExpectations]

I asked a resident of mine how he would treat a challenging case, his answer was spot on. I congratulated him. He said, he asked AI for the answer.
🤣

It is important. It is no longer about independent intelligence..

 

[Ray D. Ayshun]

When AI wants us dead in the next few weeks to months we'll have to go back to thinking for ourselves. nbd

 

[TheWallnerus]

You're someone I would hope would speak and give input.

Too pollyanna for any of it to be true imho

Check back in with each other in about 15y

 

[Ray D. Ayshun]

Look, I've seen Terminator. It's basically a cyber knife with ankles. As long as we don't give cyber knives the ability to move on uneven terrain we'll be okay.

 

[RickyScott]

ai will democratize knowledge/contouring just as the internet did 20 years ago. When I started residency, learning was through textbooks and Astro tapes and hence your program really mattered unless you made an effort to seek out knowledge. There was enormous variability in just contouring a prostate.

Not hard to see one doctor w/ pa managing 60+ pts across multiple centers in 10 years. No reason it couldn’t be done now, but specialty culture is slow to change.

 

[medgator]

ai will democratize knowledge/contouring just as the internet did 20 years ago. When I started residency, learning was through textbooks and Astro tapes and hence your program really mattered unless you made an effort to seek out knowledge. There was enormous variability in just contouring a prostate.

Not hard to see one doctor w/ pa managing 60+ pts across multiple centers in 10 years. No reason it couldn’t be done now, but specialty culture is slow to change.

Good thing we expanded residencies since two decades ago 🙄

 

[ramsesthenice]

Look, I've seen Terminator. It's basically a cyber knife with ankles. As long as we don't give cyber knives the ability to move on uneven terrain we'll be okay.

Terminator jokes became way less funny when relatively infantile AI systems learned to blackmail programmers when they felt threatened. The general concept that intelligent programs will first and foremost seek self preservation appears to be true.

 

[Ray D. Ayshun]

Terminator jokes became way less funny when relatively infantile AI systems learned to blackmail programmers when they felt threatened. The general concept that intelligent programs will first and foremost seek self preservation appears to be true.

So we'll be needed soon to replace the AI that will soon replace us.

 

[Gfunk6]

A few good uses of AI in "little" Rad Onc:

1. AI-based auto-segmentation (done by multiple companies). OARs are already quite robust, GTV/CTVs under development.
2. AI-based treatment planning. Some companies claim that ~60% can be done fully automated, especially breast/prostate new plans.
3. AI-based appeal letters on insurnace company denials.
4. AI-based physics plan-specific QA.

All of the above can be leveraged to reduce staffing now and in the future.

 

[RadOncMegatron]

A few good uses of AI in "little" Rad Onc:

1. AI-based auto-segmentation (done by multiple companies). OARs are already quite robust, GTV/CTVs under development.
2. AI-based treatment planning. Some companies claim that ~60% can be done fully automated, especially breast/prostate new plans.
3. AI-based appeal letters on insurnace company denials.
4. AI-based physics plan-specific QA.

All of the above can be leveraged to reduce staffing now and in the future.

Great list. I think it is inevitable that AI will replace dosimetrists. They probably will be the first to go, then us (rad onc might just be some random fellowship one day), then physics, then therapists.

 

[communitydoc13]

could empower smaller, resource-limited practices to demonstrate equality in care and decrease need of sending patients to larger centers. Acute attention and leverage of validated AI tools may provide comparable outcomes without necessitating referrals to high-cost tertiary centers.

This demonstration would presumably be through an accreditation program?

Do we know that quality of care is worse? For a given small center? Of course not.

How is an AI tool going to provide comparable outcomes where comparable people were not providing such comparable outcomes? Is this in fact based on the assumption that smaller centers are not staffed by "comparable people"? I'm guessing that AI "levels the field" through the phenomenon of all people doing very little. At this point, the very concept of the professional is in jeopardy.

All of the above can be leveraged to reduce staffing now and in the future.

Yes. This is almost universal. AI will reduce staffing needs in all sectors. This is in fact terrifying. At some point we will be a society of a few connected transhumanist overlords playing games with a general populace that provides value only as a consumer (or as security).

To date, I have yet to see AI "think". Now I am no expert and perhaps it does, but it seems to present things that are statistically "reasonable" based on enormous training sets. This is itself very valuable. But, I do not think that it excludes the deductive or the emotionally intelligent human.

For instance, can AI come up with something like the "principle" of least action and then be able to explain it? I'll need an example.

AI can be bullied and manipulated, this is obvious.

I have heard that our present economy is prepped up by massive AI data center and other AI investment (as consumer spending and employment stagnate). This is an enormously costly endeavor (environmentally as well as dollars wise).

Is your life notably better today due to AI?

 

[Gfunk6]

Is your life notably better today due to AI?

Yes, big time. Auto contouring alone has saved me countless hours of drudgery. Appealing insurance denials by documentation now takes seconds. Also, for private practice Rad Onc the economic realities are terminal. Less reimbursement each year coupled with greater costs. Something has to give.

For the forseeable future there will be humans at the top of Dosimetry and Physics but they will be experienced and talented, not entry level.

You look at the criteria for ACR and APEX certification and wonder why they are still living in 1990 in terms of staffing. Do you think that your lilac based SBRT is better because the Physicist is standing there?

 

[TheWallnerus]

Do you think that your lilac based SBRT is better because the Physicist is standing there?

I think I will start calling a physicist standing at the machine for SBRT “lilac based SBRT.” Flowers for Algernon!

 

[communitydoc13]

You look at the criteria for ACR and APEX certification and wonder why they are still living in 1990 in terms of staffing. Do you think that your lilac based SBRT is better because the Physicist is standing there?

Not specific to AI, but I get you're point. Staffing levels for certification need changing. But is the original poster proposing something like an AI product in-lieu of accreditation? Sorry...this sounds like grift to me. Lots of folks desperate to ride this AI bubble to their own apocalyptic bunker IMO.

"With periodically updated and ACR endorsed Safe-RT AI you will be automatically audited with real-time improvements implemented to meet the knew AI "safe standard of care" at all times". All you need to do is purchase q6 month updates with an initial investment of 300k.

I'd rather employ young, smart millennials so they can buy homes and have children.

I would personally like marked AI regulation... but I have lost faith in everything.

Regarding denials, I believe the lack of human values is what's gotten us here and AI has made it worse (but yes AI (if you call it that) is great for all things boiler plate, including insurance appeals. It'll be AI talking to AI...stoopid.).

I personally believe that AI is increasing denials, and I believe that the reluctance of human "medical directors" to challenge AI formulated payment indications has made P2P farcical. I believe that I may be seeing this in real time.

Apparently, United markedly increased denials upon institution of AI in claims processes. This all happened pre-Luigi.

lilac based SBRT

Yes, we should start giving out smoothies, facials and flowers to our patient's. This will be the differential value of the future. (CTCA sort of knew this a while ago).

 

[metallica81788]

I still don't see how AI can help my job an excessive amount
It can't hire me a new therapist in my rural area if one leaves
It can't make openings in my schedule to see new patients or assume care for older follow ups
It can't see my OTVs
It can't arrange transportation or be a social worker for my complex patients
It can't get me new referrals
It can't reduce my commute (I know that's my personal choice to have a long commute)
It can't solve reimbursement problems

I guess it could help in making decisions on challenging cases, but it can't explain that to the patients
Maybe it could help doing my notes easier
Maybe it could make my contouring of target volumes go a little faster, but I'm already using generated normals
Maybe it could make my plan review faster instead of direct communication with dosimetry

I'd be happy for somebody to convince me - but if AI is destined to improve the life of a solo semi-rural Radonc to let me make more money and have 45+ under treatment with the same effort of 20-25, I'm not seeing it

 

[drowsy12]

I still don't see how AI can help my job an excessive amount
It can't hire me a new therapist in my rural area if one leaves
It can't make openings in my schedule to see new patients or assume care for older follow ups
It can't see my OTVs
It can't arrange transportation or be a social worker for my complex patients
It can't get me new referrals
It can't reduce my commute (I know that's my personal choice to have a long commute)
It can't solve reimbursement problems

I guess it could help in making decisions on challenging cases, but it can't explain that to the patients
Maybe it could help doing my notes easier
Maybe it could make my contouring of target volumes go a little faster, but I'm already using generated normals
Maybe it could make my plan review faster instead of direct communication with dosimetry

I'd be happy for somebody to convince me - but if AI is destined to improve the life of a solo semi-rural Radonc to let me make more money and have 45+ under treatment with the same effort of 20-25, I'm not seeing it

Bridge oncology says they can fix all your problems

 

[Mandelin Rain]

I still don't see how AI can help my job an excessive amount
It can't hire me a new therapist in my rural area if one leaves
It can't make openings in my schedule to see new patients or assume care for older follow ups
It can't see my OTVs
It can't arrange transportation or be a social worker for my complex patients
It can't get me new referrals
It can't reduce my commute (I know that's my personal choice to have a long commute)
It can't solve reimbursement problems

I guess it could help in making decisions on challenging cases, but it can't explain that to the patients
Maybe it could help doing my notes easier
Maybe it could make my contouring of target volumes go a little faster, but I'm already using generated normals
Maybe it could make my plan review faster instead of direct communication with dosimetry

I'd be happy for somebody to convince me - but if AI is destined to improve the life of a solo semi-rural Radonc to let me make more money and have 45+ under treatment with the same effort of 20-25, I'm not seeing it

AI requires energy hungry data centers that raise energy prices and pollution (Beautiful coal. Burn baby burn!). The pollution directly contributes to cancer risk in your community. The increase in energy prices force people to eat cheaper, worse food and forego routine health management care raising the risk for cancer. Perhaps, also forces people to turn to vices of despair like alcohol and tobacco, further raising the cancer risk.

So you see, AI will increase your referrals. You will benefit financially as long as you have Right Think according to Peter Thiel.

 

[radiaterMike]

@metallica81788 --- I am not sure how AI will help you either. But you might only need a new therapist if they all leave.

 

[BobbyHeenan]

Radformation has helped a lot with contouring efficiency. Planning software definitely keeps getting faster too, we have to hire less dosimetrists.

Outside of that I'm not sure it's sped me up. In clinic it may have slowed me down, as patients are coming with chat GPT print outs to discuss with me.

If I were a new grad dosimetrist I would be very concerned about job market future.

 

[RickyScott]

AI is great at helping with quality metrics, lengthening notes and populating emr with statements like I counseled pt to stop smoking.

 

[OTN]

If I were an early-career radonc I can see how AI could help with efficiency. However, with more than 15 years now under my belt running a very busy service without any residents, I am so dialed in and efficient that there's no productivity gain to be had for me personally from AI. Dosimetry and physics have seen benefit and will continue to do so, of course.

In my experience AI-generated or -assisted notes are so worthless they might not as well have been done, but maybe that's just the particular product that I've seen. I use my notes to remind myself of certain things when patients enter the treatment planning process, and not having that would slow me down overall.

 

[deleted1217978]

with statements like I counseled pt to stop smoking.

But did you tho? Did you?

I have never successfully counseled a patient to stop smoking. Ever. I did convince one to stop eating a diet entirely of McDonald's tho.

 

[medgator]

But did you tho? Did you?

I have never successfully counseled a patient to stop smoking. Ever. I did convince one to stop eating a diet entirely of McDonald's tho.

It has with some of my stage 3 lungs. But you're probably right for many of those patients who say they will quit. A lot of them relapse.

It been be persuasive when I tell them it makes their treatment less effective and then more likely to lose the window of cure if they smoke during and after treatment