Beacon of Truth

Radiology nerd. ✝️ 🇺🇸🇪🇬. AI 🤖 🟪🟨 Lakers 🟨🟪

[Mark Cuban @mcuban]

No chance radiologists get replaced. A model is outdated the minute it is released.

Model usage costs money. A lot more than a radiologist.

Medicine is a business.

Putting together AI with domain knowledge is complex. When it can be the difference between life and death, who do you want building it and using it ?

People forget that using AI and trying to get accurate answers is a skill that will need to be taught in med school

AI won't replace radiologists. But every radiologist will use AI

[Beacon of Truth @radsnerd]

@cremieuxrecueil Couldn’t this just be a proxy for how expensive different place are to live in… obviously expenditure per capita will be multiple factors higher in the US, when rent is also multiple factors higher than a developing country…

[Beacon of Truth @radsnerd]

@marthanoir39 @BrianAtlas Probably yes, that’s how weight loss works lol

[Devan Sinha @DevanSinha]

I am one radiologist for one that has given up fighting enshittification & welcome the huge new revenue stream of unnecessary scanning, follow up reads & AI supervision heading my way.

Go midjourney. Let the people have as many ways to spend money on radiology as they want.

[Beacon of Truth @radsnerd]

@Evan_ss6 he must be asking so he can short it,

[Beacon of Truth @radsnerd]

@jachiam0 More data is good sure… but it’s not free. The question is what is the best use of our limited resources.

Do you think we should get daily MR scans in case anything changes? Technically more data and will lead to a quicker diagnosis if something changes

[Beacon of Truth @radsnerd]

@joejanizek Mind you these are the pictures they chose to include… this is the best they could come up with

[Beacon of Truth @radsnerd]

@radfugee A slightly different opinion- if all of their goals go according to plan, it will be a decently sensitive but non specific whole body scan; meaning a ton of healthy follow up CT/MRs… which sounds like easy RVUs… maybe a win win? Some of those incidentals may be early malignancy

[Beacon of Truth @radsnerd]

@WillyRontgen @francisdeng @cookyscan1 @VChernyakMD @heacockmd If all their goals go according to play, and they are wildly successful, all it means is that radiologists will have to read a ton of healthy CT/MRs to follow incidental findings… continue exploding imaging volumes but at least it’ll be easy scans and RVUs

[Beacon of Truth @radsnerd]

@Honest_Abe_1864 @septisum 🤦‍♂️

[Logan Bolton @ EMNLP 2026 @septisum]

His reasoning for those curious:

“Most recent efforts at whole body imaging screening have been whole body mri. The incidental findings for each patient can run hundreds of thousands of dollars, last for years, and lead to innumerable unnecessary biopsies, surgeries and complications.

There are some definite screening interventions we can take, but most require preventing damage (blood pressure and lipid control for blood vessels) at a microscopic level or extremely early detection of cancers (mucosal lesions on colonoscopy, subtle lesions on breast mammograms or breast ultrasounds/ breast MRIs). Almost anything you detect on these imaging scans will be damage already done or metastatic disease and will result in a lead time bias with no true evidence of decreased morbidity or mortality.

The physics for doing this with ultrasound are also against them. For proper deep tissue entrance and reasonable scan times, even with “advanced tech”, the energy of ultrasound source has to be high enough that they will be limited by tissues getting hot. Even if the whole thing is a failure, there may be some important technical developments that come out of the ashes.”

[Beacon of Truth @radsnerd]

@aykutuz @grok @FredLambert What I mean is that getting models to interpret scans we already have infinite datapoints on is difficult but is the way to move forward. New hardware modality is a separate issue. 3D US is not a new idea and has physics based limitations. It doesn’t have much to do w/ AI…

[Beacon of Truth @radsnerd]

@aykutuz @grok @FredLambert You seem like a smart guy but you’re conflating a new imaging modality (hardware) with interpretation. Radiologists interpret US, XR, CT and MR. Getting software to segment and interpret scans is not even close to being solved and akin to solving full self driving in cars.

[Beacon of Truth @radsnerd]

@heacockmd @DanielSodickson @NYUImaging Most likely what will happen is we will get a lot of healthy people scanned with sub optimal scans, lots and lots of incidentals will be picked up; these people will then go get medical scans through the proper channels to tell them about their benign hemangiomas and lipomas

[Beacon of Truth @radsnerd]

@heacockmd @DanielSodickson @NYUImaging Unless AI is able to change physics such that ultrasound can penetrate bone and air, color me very skeptical 🤨

[Laura Heacock, MD @heacockmd]

Reposting this because I think it eloquently stated the same points I made today re: this new product vs MRI, and also because it includes the thoughts of Function’s @DanielSodickson, formerly of @NYUImaging and truly the expert’s expert on MRI.

[Wil @WillyRontgen]

Detecting a small renal cancer in 2026. CT/MR/Midjourney. What are you choosing?

[VFP le scintigraphiste @Scintigraphiste]

À gauche, les images révolutionnaires qu'ils annoncent. À droite un scanner banal.

[Beacon of Truth @radsnerd]

@martinez_clips 🤣🤣🤣🤣 everyone in the comments can see how retarded you are

You could be arguing the most sensible thing in the world and would find a way to butcher it like a jackass

[Beacon of Truth @radsnerd]

@riqposso @martinez_clips 100%, the only annoying thing about this guy is how he’s convinced he’s smarter when it’s clear to everyone else he’s retarded 😂😂😂

Martinez buddy everyone is laughing at you