It's pretty wild you're telling a practicing radiologist how they should be doing their job.

One of the problems with letting people in a profession set the requirements for that profession is that they exaggerate them to limit competition and increase their earnings. You end up with things like radiologists saying they need to know cellular biology. It's simply ludicrous on its face.

Radiology is a particularly egregious case. If it weren't for legal protection the profession would be dramatically different. It's obviously stupid to spend a bunch of time training them to interpret and diagnosis a bunch of different scans. Basic efficiency would be to train Person A to interpret chest X-Rays, person B to do CT scans, Person C for leg X-rays and have them spend all day doing that one thing. Instead, radiologists spend a decade learning how to interpret a hundred different things and spend 1/100 of their time on each of them.

How do you interpret an image if you don't have an understanding of what is in it?

In other words, let's say that a patient has a fever. Chest X-ray was done.

High-school educated "radiologist" sees the image. Does he understand the anatomy? The variations of anatomy? Pathological manifestations of potential causes of fever? How to exclude image artifact versus include potential sign of pathology? What about findings that are not related to fever but need to be identified, further characterised and further imaging required for follow-up? What about signs of infectious fluid versus non-infectious fluid like blood or extravasated fluid?

What about when the ordering physician wants to discuss the findings with the "radiologist"? Will that "radiologist" actually understand anything that the physician is talking about?

There is a role for AI assisting in rapid analysis of radiological studies, however radiologists can never be replaced because AI will never perform to the level of or have the same functions as a physician radiologist.

>How do you interpret an image if you don't have an understanding of what is in it?

Because you're looking to identify patterns and match them to known ones. To use an analogy, I can teach you to identify statues of Hindu Gods without teaching you anything about Hinduism. For example, to identify Ganesh you need to know that he has an elephant head. Knowing that he has an elephant head because Shiva cut his human head off doesn't really help you.

>In other words, let's say that a patient has a fever. Chest X-ray was done.

>High-school educated "radiologist" sees the image. Does he understand the anatomy? The variations of anatomy? Pathological manifestations of potential causes of fever? How to exclude image artifact versus include potential sign of pathology? What about findings that are not related to fever but need to be identified, further characterised and further imaging required for follow-up? What about signs of infectious fluid versus non-infectious fluid like blood or extravasated fluid?

All of the things you mentioned are things that a radiologist interpreting a chest X-ray needs to know. What you need to explain is why someone interpreting a chest X-ray needs to understand Organic Chemistry. And the vein structure of the leg. And the typical development pattern of a child. And the various mental illnesses a person might have. And interpret a knee MRI.

>There is a role for AI assisting in rapid analysis of radiological studies, however radiologists can never be replaced because AI will never perform to the level of or have the same functions as a physician radiologist.

No, but I could easily take 10 people, train them each 1/10th of what a radiologist studies, and have them perform just as well as 10 radiologists by routing the right stuff to the right person.

The average Radiologist does not only interpret a single modality or the same body part day in and day out (Radiograph, CT, MR, ultrasound, nuclear scintigraphy, PET, or mammography). We are prepared by our extensive residency to competently read any and all of these upon graduation. We do not interpret these in isolation; we are constantly comparing to prior studies across modalities, often across overlapping body regions. I completely disagree that barely trained hyperspecialists would be sufficient.

While specialization is coming to all of Medicine, we tend to cluster by body region/disease state, rather than by modality. This is because all of the modalities provide complementary information and you must be able to cross reference across the various manifestations of disease.

All of medical knowledge is iterative. Unlike programming, you can’t just abstract the low-level programming. The premedical curriculum provides the baseline knowledge to understand pharmacology, which is essential when trying to understand our interventions in physiology and pathophysiology. We need to know both physiology and pathophys when interpreting imaging to know how disease manifests and what is abnormal versus post therapy related change.

We have to know how referring clinicians will treat disease, and know the major complications to look out for on imaging. We discuss their treatment plans during tumor boards and need to speak the language of the treating teams so we can tailor our interpretations to be useful.

There are no easy shortcuts here. NPs and PAs are a living experiment at a shortcut, but what I see day in and day out is that the people they consult (radiology and pathology) need to know even more clinical medicine to help the inexperienced NP or PA in knowing what to do when something happens that deviates from the protocol. Many many many times I will call with a semi-urgent unexpected finding, and just get silence on the other end of the phone. They don’t know what to do, whereas on weekends or nights when I get residents or attendings, I don’t hear this complete absence of understanding.

> I completely disagree that barely trained hyperspecialists would be sufficient.

>While specialization is coming to all of Medicine, we tend to cluster by body region/disease state, rather than by modality.

These are contradictory statements. If Doctors typically go:

No Training -> Trained in everything -> Specialized in body region/disease state

You can easily go:

No Training -> Trained in body region/disease state -> Specialized in body region/disease state

So even if we assume it's not possible to specialize radiologists any further than they already are we can still cut the training time and difficulty.

>We have to know how referring clinicians will treat disease, and know the major complications to look out for on imaging. We discuss their treatment plans during tumor boards and need to speak the language of the treating teams so we can tailor our interpretations to be useful. We need to know both physiology and pathophys when interpreting imaging to know how disease manifests and what is abnormal versus post therapy related change.

No one is denying that there are things in medicine that are complicated and require very skilled people. What you describe is the end state of a fully educated and experienced doctor. Medicine is more or less the only field that makes you become that before you start working. Can you take a reasonably smart person off the street and have him designing plans to treat tumors in a year? No. Can you take that person and train them to identify normal appearing lungs vs cancerous ones in that time? Probably. And can that person learn on the job and develop the ability to design a treatment plan over the course of 10 years working? For sure. And they wouldn't learn everything taught in Med school. They would pick up only the things relevant to the job they are trying to perform.

>All of medical knowledge is iterative. Unlike programming, you can’t just abstract the low-level programming. The premedical curriculum provides the baseline knowledge to understand pharmacology, which is essential when trying to understand our interventions in physiology and pathophysiology. We need to know both physiology and pathophys when interpreting imaging to know how disease manifests and what is abnormal versus post therapy related change.

Going back to the comment that kicked this all off. If I handed you a cellular biology final from Med school do you think you'd be able to pass it? What about the USMLE?

>There are no easy shortcuts here. NPs and PAs are a living experiment at a shortcut, but what I see day in and day out is that the people they consult (radiology and pathology) need to know even more clinical medicine to help the inexperienced NP or PA in knowing what to do when something happens that deviates from the protocol. Many many many times I will call with a semi-urgent unexpected finding, and just get silence on the other end of the phone. They don’t know what to do, whereas on weekends or nights when I get residents or attendings, I don’t hear this complete absence of understanding.

Every other field has the ability to take inexperienced people and make them experienced. There's no reason medicine can't do the same. Your clueless NP and PA should have more experienced people to go to for help when they experience something new. The next time they get that call they wouldn't be as clueless.

What you are describing is literally how medical training works. Residency IS on-the-job training. Medical school provides the foundation to then learn on-the-job; residency transitions them with graded responsibilities to independent practice.

As far as your targeted questions, yes, I do think I would pass a cellular biology test from medical school. I wouldn't get 95%+ but I would expect myself to get at least 80% correct.

I'd estimate the overwhelming majority of practicing physicians would pass the USMLE if they took it cold. They probably wouldn't do as well as they did after many weeks of dedicated study, but they would be above the minimum threshold.

>What you are describing is literally how medical training works. Residency IS on-the-job training. Medical school provides the foundation to then learn on-the-job; residency transitions them with graded responsibilities to independent practice.

It's not because you can't start on the path without going to medical school. And you don't need medical school to start learning radiology. I messed up my elbow in high school. They showed me the x-ray and said this is a fracture, this is a chip, etc. It wasn't difficult to understand and with enough time and practice I'd be able to see them in other x-rays. I don't see where something like Organic Chemistry comes into play.

>As far as your targeted questions, yes, I do think I would pass a cellular biology test from medical school. I wouldn't get 95%+ but I would expect myself to get at least 80% correct.

So ~20% of what you learned in cellular biology isn't needed as a radiologist assuming you only remember what you've actually needed.

There are a thousand reasons why your theories make absolutely no sense at all in the real world. You also make baseless and downright false assumptions due to your complete ignorance. People like you are part of the issues that make the job harder than it should be. Maybe you could take a break and re evaluate whether you have the elements to make judgement calls? Or enter med school and see if your theories hold water?

You “could” teach a high school student all aspects of medicine, but it’s going to take so long for them to build the groundwork that by the time they can meaningfully contribute, you’d never make the return on investment. This is actually how medical training was done in the years prior to the Flexner report.

From what I can tell in this exchange is that you have an innate distrust of credentialing, because you think because there is some excess therefore the entire system is wasteful. I’d argue that this is the DunningKruger effect in action.

And claiming that 20% of cellular bio is unnecessary is a silly metric when in the US 70% is passing... I don’t think that logical conclusion applies.

Yes we have more tools to look up information that we have forgotten, but as another poster here says “we aren’t bootstrapping the search each time”. If there’s something we’ve forgotten, it’s much more effective knowing where to start the search.

I agree there’s a lot wrong with our current training paradigm; the cost of medical school education is #1 with several terrible downstream effects.

However, there are no shortcuts. You have to put in the time to master the material, else you are doing your patients a disservice.

> It's simply ludicrous on its face.

That's not a meaningful test of correctness.

Many medical practices are organised as you describe (for example, surgeries specialising in nothing but hernias), but medical cases often transcend simple boundaries and a broad amount of understanding is required to detect that boundary violation (such as realising it's not a hernia, or that an organ near the hernia seems to be diseased, or that the patient is going into anaphylactic shock because of a surprise reaction to anaesthetics, or ... or ... or ... or ... ).

> or that the patient is going into anaphylactic shock because of a surprise reaction to anaesthetics, or ... or ... or ... or ... ).

And if I go into anaphylactic shock during my hernia surgery should I be grateful that my surgeon did a Pediatric, OBGYN, Neurology, Psychology, and Oncology rotation? Or that they can draw the Krebs cycle?

The level of aggressive ignorance displayed in comments like this is quite shocking. I'm puzzled as to why engineers and developers think they're qualified to give advice to other professionals in fields where they have no education or experience. What causes that type of hubris?