Why You Should Do Diagnostic Radiology

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

talkingcheese

Full Member
10+ Year Member
Joined
Sep 29, 2010
Messages
14
Reaction score
21
I posted this on reddit/r/medicalschool a few months ago but feel like this is when MS3 start to make decisions about specialties. Hope this helps!

Background: I’m a chief resident, PGY-5, at a mid tier academic program in a big city. Traditional route med student who didn’t know they wanted to do rads until the beginning of 3rd year. Love the field and think there’s a lot of misconception among med students of what it entails.

Radiology years:


PGY-1: Intern year- can do a preliminary medicine, surgery or transitional year. Do the easiest thing you can, and if it’s in the same city as your advanced, sweet.

PGY-2: Radiology R1- The ACGME says 1st year residents can’t take call, making this possible the easiest year on your entire training hours wise. You will be overwhelmed by a completely new way of looking at medicine, but you won’t work nights or any (or very many) weekends. You should be studying to prepare you for….

PGY-3: Radiology R2- The hardest year of residency work wise. Very call heavy. I worked almost 3 months of night float, and more weekends than I can count. However, this is still better than what people in a surgical or medicine residency go through!

PGY-4: Radiology R3- Boards. So radiology does this weird thing where you take a monstrous CORE exam at the end of this year, which is actually only a “Board Certifying exam”. It’s incredibly difficult, requires months of preparation, with a 80-90% pass rate (but remember, this is a group of nerds with a Step average of 240+ you compete against). There’s an annoying physics section. Time “off” to study is variable per program, but you will be studying several hours a day starting in the winter. The actual “radiology boards” is taken 18 months after residency (during your actual job), and is a joke with close to 100% pass rate. Most programs also allow residents to go to Washington DC for a month (paid for) for AIRP, a radiology pathology lecture course. It’s like being a med student with no responsibilities and no tests for a month. Great for those who don’t have a family to leave behind.
You will also apply to fellowship this year. Most fellowships are going towards a match (except Body, Chest), and you will apply and interview second half of the year.. Annoyingly during boards prep time. You also find out where you match a couple days after you take boards. Vast majority of fellowships are 1 year, and everyone does one.

PGY-5: Radiology R4- Boards behind you and majority of call behind you. Most programs let you dictate your schedule, with “mini fellowships” of 4-6 months in your subspecialty of choice. Usually in something to compliment what your actual fellowship is in.



Typical day:



An example of a typical day of a resident on a diagnostic rotation, such as Musculoskeletal.

8:00 AM Arrive and start “dictating” the studies on the list, which would be plain extremity x-rays or MRIs (knees, shoulders etc) depending on your seniority. This involves dictating a preliminary report of the study, that won’t go out until the attending reviews it.

9:00 AM Go perform a joint injection (fluoroscopic guided hip or shoulder injections mostly, for steroid and pre-MRI arthrogram). These occur anywhere from 2-5x a day, and usually take 15-30 minutes depending how fast you are. You do all the set up and the attending will come to watch when you’re actually injecting.

10:30-Noon Go “sign out” the studies (xrays, MRIs) you’ve read with the attending. This is usually sitting next to them while they look over the study and your reports, going over findings and occasionally pimping you.

Noon-1pm: Conference, half are pure didactic and half are case based. Radiology makes it really easy to have a hundred images of different pathologies and going around the room having residents work through them/answer. Radiology is also much heavier on conferences than other specialties, averaging around 5 hours a week in most programs. This will go way up for R3s during board studying time.

1pm-5pm: Repeat of the morning



Other diagnostic rotations would include Neuroradiology, Ultrasound, Body CT, MRI, Breast etc and they all have their own procedures including lumbar punctures, myelograms, thyroid biopsies, breast biopsies etc. There are a lot of procedures outside IR, something I wasn’t aware of before residency. This varies by institution however.



Call: Overall hours wise as a resident you will not be there that long unless call is involved. Call schedules vary so much among programs that saying mine won’t really help, but R2 year is the busiest with around 1-2 months of night float, and 10-15 weekend day coverage. Radiology doesn’t have separate residents on for different specialties for call (such as a MICU night float, cardiology night float etc) so at most you will have 1-2 residents in the hospital covering ANYTHING radiology related. This includes for us

1. Dictating every study done on hospital inpatients (minus ICU chest xrays), everything coming through the ED & multiple satellite urgent cares. A car crash with 4 passengers at once? You’ll have 4 CT Chest abdomen pelvis, CT Heads, CT c-spines and an xray of every extremity that hurts on your list at once, with the ED calling you asking for results. It’s overwhelming and exhilarating (for some)
2. Answer calls/pages for anything radiology related, including review studies with surgeons on call, questions about what to order etc.
3. Performing any diagnostic radiology procedures, including: septic joint aspirations that need fluoroscopy (hips), fluoroscopy guided lumbar punctures, esophagrams for perforations, intussusception reductions.
4. Fielding IR consults, meaning gathering all the info, consenting, calling in the team and IR attending to perform it. Sometimes we scrub in on these but usually the diagnostic part is so busy we can’t. Some programs with bigger IR sections will have fellows on call to handle this.

Call as a resident is always in house, you will probably never sleep. It is very different that how other specialties handle call. A busier call lets us enjoy a lighter regular schedule.



I love radiology as a field, and try to convince every med student to do it. Here’s some reasons why:



Pure medicine, no BS: I believe the 2 big reasons someone pursues medicine are the humanitarian aspect and the science aspect. I leaned towards the latter, and most people I’ve encountered in radiology are the same. My biggest gripe about intern year was how little medicine you do. Pretty much all the data gathering and analysis, including differential for a service could be done in an hour, but you spend the remaining 12 hours calling consults to regurgitate information, call social services, and essentially act as a secretary. This obviously reduces are you become more senior, but hospitalists still do this. Radiology is just you and a study, trying to get information out of it. Non-compliant patient with crazy abscess? I diagnosed it on CT in 5 minutes and told the clinician, now it’s their problem trying to get him to take antibiotics. Diabetic with necrotic pancreatitis? I diagnosed it on CT in 5 minutes, and wash my hands. You can help so many people, so fast, because you are just doing medicine. Never have to deal with insurance issues, getting someone in a nursing home, trying to get a consult to see a patient. Never have to deal with getting “dumped on” at 5pm, because even if a MRI comes on the list at 4:59? I’ll read it in 10-15 minutes, vs an hour admitting a new patient.

This gets me to the next point: You control your own pace. There are no nurses you are waiting to get labs, no attending sleeping at home you’re trying to get to round (we do have attendings that read out studies slower than others, but magnitudes less painful than rounds, and this disappears when you’re an attending yourself). You don’t have to wait for pancreatitis to resolve to discharge a patient, you just sign the study and you’re on the next one. Never have a million checkboxes to do for the day, just clicking on study at a time (with procedures thrown in).

Interacting with colleagues more than patients: If you love patient interactions, radiology is probably not for you. However, you can still be plenty social in radiology. The only difference is you just spend the entire day talking to coworkers (who are in the same rooms as you) and a bunch of consulting clinicians. I really enjoy talking to clinicians about studies and reviewing them, as opposed to a patient who doesn’t know anything about their care. A perfect medium would be the ability to just review studies with patients who are interested, but don’t think that’s going to be a billable code anytime soon.

Finally: It’s like learning a new language. Every service thinks they can read their own studies (and some can), but vast majority of clinicians have no clue beyond a basic xray. I still remember telling a pulmonology fellow, when I was an R1, that the pneumothorax he was worried about was just a skin fold. Even early in your training, your abilities will surpass that of attendings in other fields and it feels… awesome. To have someone call with a study saying “I have no idea whats going on” and you know what it is 2 seconds in, is a great feeling. This feeling will only get exemplified in private practice, where everyone is more reliant on radiology.



So how do you know if radiology is right for you? Here a some characteristics I think that may be a sign that it’s for you.

You liked the first 2 years of med school more than the 3rd year. I was miserable 3rd year, because most specialties have the social work mentioned above, and you’re never in control of your own time. Even intern year was better, but I didn’t really ever enjoy coming to work till I started radiology residency. I have a great social life, but I really enjoy coming to work and just having it be me and my work, with some interactions with colleagues. It’s very similar to studying a bunch the first 2 years (with more free time).

You are organized/efficient. The people I see struggling most in radiology are those who are slow. This does not mean they make bad radiologist, on the contrary, I would want a slow methodical radiologist to read my scans. But to be able to keep up with the pace that the field demands and enjoy it, you have to have some sense of speed. Being a techy is also related somewhat, but we’re definitely a minority even in radiology.

You do not have an ego. Radiologist will get **** on by every speciality, “clinically correlate” etc etc, and some specialties such as ortho or neurosurgery will pretend you don’t exist. But you have to be ok with not being in the front seat of patient care, and making contributions behind the scenes.

You are a good test taker. Radiology is essentially one big test. Staring at the screen, coming to an answer. This will also help with the CORE exam..



Dismissing some misconceptions about radiology:

AI- I’ve literally never heard a radiologist bring this up as a viable threat. We are 20+ years from this making any significant impact, and when it does, it’ll just make our lives easier. We’ll have a 50% unemployment rate from machines before radiologists are actually put out of jobs. Do not worry about it.

Outsourcing- Also not a real threat. Clinicians want to talk to their radiologists, which is why not every radiologist is a teleradiologist. There’s a handful than get US board certified and go overseas to read US studies, but this is so rare that it’s a non-issue (and doesn’t save that much money). Teleradiology is definitely a big thing (reading studies from states away), but is seen as a last resort by most due to poor compensation.



Some real downsides to the field:

You will work hard. Attending radiology is not a cush 40 hour work week. This exists in some settings (VA especially), but most are pushing 50-60 hours, with 10+ hour days. These days are BUSY, reading studies and doing procedures non-stop. If you want to have a lot of downtime at work, radiology is not for you. To make up for this, most private practices offer 8-12 weeks of vacation, which can only happen because we have no continuity of care to worry about. Working harder for the same amount of pay is universal in medicine however.

You always have to be “on”. You can have a bad day as a hospitalist, maybe half-ass some physical exams and be ok, but if you half-ass some studies, I guarantee you’ll hear about the cancer you missed on the chest x-ray in a few years. Majority of my misses as a resident have been when I’ve been pushing myself to read faster than I should, or was in a hurry to finish. Radiology is unforgiving.

Attending life is harder than resident life. As above, your hours get worse (no nights though, that's usually taken care of a hired nighthawk service) and days more stressful because of all the litigation risk, but the pay and vacation are there for that. I’m sure med students are very interested in pay, but I don’t have information that can’t be found online (see doximity compensation report). Of note, the regional variation is huge and you can make family med money in downtown of a big city vs surgical subspecialty money by going rural.



Hope that helps. I feel like the whole application process and score averages have changed since I applied so not sure how much help I can be of that, but some things: Step 1 is big, research isn’t really (I had nothing). The tier of program only matters if you want to do academics, location is way more important for connections.

Members don't see this ad.
 
  • Like
Reactions: 20 users
Couple questions:

1. I always hear that get to do procedures in residency but how many exactly? Is it one procedure the entire day? One every two hours? I figure it depends on what service your on but I’m curious how often are you leaving the reading room to do that?

2. When your on a service like IR for the month, how much of the procedure do you get to do? Am I just assisting the attending or can I actually be controlling the guidewire, catheters, etc and the attending watches?

3. As an attending, do part time radiology jobs exist (I don’t care about the pay cut) or will groups not hire you at all? I don’t mind working 50-60 hours a week but I’m curious to know if its actually possible to scale back like other specialities.

4. I’m interested in IR and wanted to know how busy is call? I figured since we are not primary, you wouldn’t be consulted that often. Do that many patients need IR services on a daily basis for call to be that busy? IR attendings always talk about coming in for GI bleeds but I never understood why GI can’t take of that with some endoclips. I’m only a MS3 so excuse my lack of knowledge in this aspect.

5. If you ending up becoming an IR attending and when you get older, don’t want to work as much, is it a viable option for you to switch over to just reading films and not doing procedures? Or would the group you work for not allow that because they have DRs for that?

6. How is the day-to-day schedule different from a DR attending in academics vs PP? Are there vast differences in hours?

7. What are my chances at Top tier, mid tier DR academic programs with ESIR?
DO student at a low tier school in northeast.
Step 1 238
Top 25% class
HP in clinical rotation grades so far (IM, Surg)
Research: 2 first-author neurosurgery publications, 1 possible first-author IR publication, possiblely an IR abstracts.
ECs: Professional Dancer
 
Last edited:
I posted this on reddit/r/medicalschool a few months ago but feel like this is when MS3 start to make decisions about specialties. Hope this helps!

Dude that's a great post! I felt I recognized myself, except for the test taking part... Really great info to share with all the students.
 
  • Like
Reactions: 1 user
Members don't see this ad :)
Hello
Does anybody have any info about AI?
It really stresses me out
I mean you go to med school and do a residency for 10+ years hoping that you have a long successful carreer (30+ years)
Why should you go into a field where this is not guaranteed?
 
3. As an attending, do part time radiology jobs exist (I don’t care about the pay cut) or will groups not hire you at all? I don’t mind working 50-60 hours a week but I’m curious to know if its actually possible to scale back like other specialities.

You could ask this question about pretty much any type of job in radiology, and the answer would be yes. So, yes, they exist. In academics, it's pretty easy to just be a 0.8 or 0.6 FTE, with proportional decrease in call. In private practice, my experience is that part-time positions are employed ones, usually with no call in addition to, for example, a 4 day work week and less vacation. You're usually taking a pretty decent paycut for this lifestyle choice, so you might be doing half the work of the partners, but you'll only get 40% of the pay. The exception is probably breast imaging, where someone willing to do that can become partners and/or make more money while still being part-time.

6. How is the day-to-day schedule different from a DR attending in academics vs PP? Are there vast differences in hours?

Hours? A little, but probably not as much as you think. I had more flexibility in my day in academics, meaning I could run an errand over lunch while the fellow covered. But, when taking call out of the equation, a normal work week was really just a few hours shorter than now. The real difference is pace. You'll read a vastly greater number of studies in private practice, and the days can easily get hectic. That lunchtime errand is a nonstarter, and some days I'm lucky to make it to the doctors' lounge to grab something and make it back to the reading room before lunch is over.
 
  • Like
Reactions: 1 user
Why should you go into a field where this is not guaranteed?

Because nothing is guaranteed? And because medical students have been perpetuating misconceptions about radiology for at least 20 years? And because, when our AI overlords come for us, radiology will be far from the only thing - in medicine or otherwise - to become automated?
 
  • Like
Reactions: 5 users
Hello
Does anybody have any info about AI?
It really stresses me out
I mean you go to med school and do a residency for 10+ years hoping that you have a long successful carreer (30+ years)
Why should you go into a field where this is not guaranteed?

Because every other specialty sucks and Radiology is the best?
 
  • Like
Reactions: 3 users
This is a brilliant thread. Well done OP. I think being a prospective applicant to radiology poses a unique challenge that doesn't exist across most other specialties. Most medical schools don't devote required curriculum time to radiology. It's not uncommon to go through an entire MS3 year without seeing a reading room. Fortunately this has been changing, as medical schools realize how integral imaging has become to medicine. Nearly every specialty is heavily reliant on radiology. On a side note, dependence on radiology will only continue to increase, something that is a huge pro for the future of the field moving forward.

The more difficult barrier to overcome as a prospective applicant though, is a developing an understanding for what the day-to-day for a radiologist is. In most of the core clinical rotations in medical school, you can gain reasonable insight into what that job feels like. In medicine clerkship you perform all the functions an independent hospitalist would, albeit on a smaller scale. In surgery clerkship you suffer through early rounds and get time in the OR (hopefully at least doing some suturing). Same thing for OBgyn, peds, psych, FM etc.

Radiology though is fundamentally harder to access. Most radiology elective experiences are comprised of shadowing. I can tell you as someone who loves radiology residency, my radiology elective was incredibly boring. After making it through MS3 without coffee, I had to take up a coffee habit to stay awake in the reading room. No joke. For all fields in general, shadowing is a poor way to gain insight, and radiology is no different.

Some schools have been moving towards a radiology "sub-internship" which I think is fantastic... actually getting to look at images and dictate your thoughts into a report. I hope this becomes more commonplace. With that said, even with a hands on experience, it's still difficult to get a feel. The learning curve in radiology is so unbelievably steep, It's difficult to gain enough competence to have some clue what you are doing.

Ultimately, most applicants windup having to take their best guess on whether radiology is the right fit. I know that's what I did. I think the OP's post hit the nail on the head in terms of what to look for. Radiology is a field jam packed with critical thinking and medicine, with a minimum amount of BS. It's a shame that it's difficult to appreciate this at the medical student level.
 
  • Like
Reactions: 4 users
Is there any bias towards DO medical students wanting to get into radiology? Have you personally worked with any DO radiologist?
 
Is there any bias towards DO medical students wanting to get into radiology? Have you personally worked with any DO radiologist?

Over all? Yes. Program directors would probably prefer to fill their classes with qualified US allopathic seniors if possible. The fewer DO's, IMG's and FMG's the more "competitive" the class appears, which has appeal for applicants.

Now does that mean as a DO student you won't match anywhere? Of course not. My low/mid-tier residency had several DO residents and there wasn't any noticeable difference between them and the MD residents. My program recognized a few DO med schools that reliably produced solid residents, like TCOM in Texas, and took a couple DO's every year.
 
  • Like
Reactions: 1 user
Is there a list of categorical programs? I will be couples matching and would prefer to not have to move or be away from significant other for a year during PGY1. Any information would be appreciated if anyone here has any.
 
Is there a list of categorical programs? I will be couples matching and would prefer to not have to move or be away from significant other for a year during PGY1. Any information would be appreciated if anyone here has any.

The NRMP results file allows you to identify categorical programs, which are designated with a C code, rather than an A code.
 
  • Like
Reactions: 1 users
Radiology is a field jam packed with critical thinking and medicine, with a minimum amount of BS

I love this concept and it's what drew me to DR. But as an MS3 having just finished IM, many of the radiology reads had a lot of nonspecific findings with no definite diagnosis which we were told to "clinically correlate." My question is, how often on a daily basis do you have the time to check clinical findings from the chart and correlate that with the read? To me, the diagnosis is the most satisfying part of medicine and I feel as if "diagnostic" radiology does not do much of the diagnosing.
 
  • Like
Reactions: 1 users
Members don't see this ad :)
Fear of artificial intelligence in radiology is equivalent to fear of non-MD human intelligence in every other medical specialty. An NP or PA makes an internist more efficient, which decreases the number of internists needed for any given number of patients. What's the difference? Take it one step further - a medical assistant takes the history and review of systems, a scribe does the documenting, a pharmacist does the medication reconciliation. The MD shows up for the exam, decision-making, and occasional procedures. What's the difference?
 
Last edited:
  • Like
Reactions: 4 users
I love this concept and it's what drew me to DR. But as an MS3 having just finished IM, many of the radiology reads had a lot of nonspecific findings with no definite diagnosis which we were told to "clinically correlate." My question is, how often on a daily basis do you have the time to check clinical findings from the chart and correlate that with the read? To me, the diagnosis is the most satisfying part of medicine and I feel as if "diagnostic" radiology does not do much of the diagnosing.

KermitNodding.gif
 
AI will make radiologists much more efficient, which will decrease the number of radiologists needed for any given amount of work.

It’s coming. If you won’t be an attending for 10 years, it might be an issue.

Then again, it will be an issue for every other non-surgical specialty soon enough too. And eventually the surgeons.

tasks.png


This is an XKCD comic from 2014 (made by a really smart former NASA engineer if you’re not familiar). Classifying something as a bird used to be really hard for a computer to do, very recently.

You can now solve this and similar questions with a few lines of code. That’s why the analogies to Mammo CAD are silly - any comparison from before 2012 doesn’t count.

We underestimate how complicated the visual tasks we find easy are thanks to evolution, and overestimate how hard more technical questions are.

We’re not used to it, but it’s easier for a computer to learn to identify a PE than a bird.

So would I recommend radiology to an MS1? Maybe if you have an MS in CS or plan to do IR. Otherwise, caveat emptor.

The fact is no one knows. You are just speculating. Medical students should do what best fits their personality and what they enjoy. I know I would be miserable in every other specialty. I'm not worried about AI, we will always figure out a way to stay relevant. Its been that way for decades. Imaging will only get better and more sensitive/specific, as we move away from invasive tests, biopsies and colonoscopies.
 
Last edited:
  • Like
Reactions: 1 user
Some of us will stay relevant, not all of us.

And there may not need to be as many of us.

Your argument works equally well for the safety of lying down on train tracks.

No matter how much you try, you can't predict the future. All speculation. MRI was supposed to shrink our field and make us irrelevant. About that....I can also speculate that the world will end in 2030 or that I'll wake up Lebron James tomorrow morning.
 
  • Like
Reactions: 1 users
I love this concept and it's what drew me to DR. But as an MS3 having just finished IM, many of the radiology reads had a lot of nonspecific findings with no definite diagnosis which we were told to "clinically correlate." My question is, how often on a daily basis do you have the time to check clinical findings from the chart and correlate that with the read? To me, the diagnosis is the most satisfying part of medicine and I feel as if "diagnostic" radiology does not do much of the diagnosing.

Quite often actually. I frequently pull up the note of the encounter where a study was ordered to get an idea of what the clinician was thinking. That being said, there's a limit to how much time you can spend going into a chart. The medical record is filled with a lot of discrepant and/or flat out inaccurate information and sometimes no amount of extra time spent will magically make the diagnosis more clear.

Besides, once you've been in radiology long enough you'll understand that a lot of very different stuff can look identical on imaging. Sulcal FLAIR hyperintensity can represent subarachnoid hemorrhage, sarcoidosis, lymphoma, metastatic cancer, meningitis or hyperoxigenation. Most of the people I trained with abhorred the phrase "clinical correlation" but sometimes it's necessary. Hopefully the report spells out what further correlation is needed, e.g. a new lab or focused physical exam, cuz otherwise its a throwaway phrase.
 
  • Like
Reactions: 1 user
I love this concept and it's what drew me to DR. But as an MS3 having just finished IM, many of the radiology reads had a lot of nonspecific findings with no definite diagnosis which we were told to "clinically correlate." My question is, how often on a daily basis do you have the time to check clinical findings from the chart and correlate that with the read? To me, the diagnosis is the most satisfying part of medicine and I feel as if "diagnostic" radiology does not do much of the diagnosing.

I feel you. I'm a resident so I have time. Every cross-sectional study warrants a glance at the chart to inform what I'm looking for, because the provided history in the requisition is usually inadequate. When it comes to reads, there is a fine line to toe: be specific to be valuable, don't be wrong, and don't step on the toes of your referrers who may reasonably disagree with your assessment based on findings not yet documented in the chart. How you navigate that line is stylistic, attending-variable, and time-dependent. Whenever I detect a potentially significant but nonspecific finding, I'll look at the chart again to see if that can be made more specific or certain. I'll describe the findings nonspecifically in the body of the report and be more specific in the impression when possible. The impression is where we get to reason like a clinician, integrating findings and chart information. The most satisfying reports involve diagnoses that are new, unsuspected, clinically important, and reasonably certain/specific. They happen more than you'd think. Not every day but maybe every week, depending on subspecialty/setting (what do other rads think?). Remember that as a radiologist you'll come across way more patients than you would as a bedside clinician, and each patient is a new opportunity to drop the power-mic on a significant diagnosis.
 
  • Like
Reactions: 1 user
Fear of artificial intelligence in radiology is equivalent to fear of non-MD human intelligence in every other medical specialty. An NP or PA makes an internist more efficient, which decreases the number of internists needed for any given number of patients. What's the difference? Take it one step further - a medical assistant takes the history and review of systems, a scribe does the documenting, a pharmacist does the medication reconciliation. The MD shows up for the exam, decision-making, and occasional procedures. What's the difference?

I agree. I shadowed a DR during the winter break and he told me during his residency (early 2000’s) he was told that AI was going to “take over” the radiology field, yet he’s been practing for 10 years plus and he hasn’t seen any difference within the field. He told me to not pay attention to any of that, if I did choose DR as a specialty in the future.
 
I die a little inside whenever I see or hear someone use "correlate clinically" in a general sense. If the person on the other end of the report doesn't already know to do this, then the radiologist telling them to do it won't help. It's one of several ways radiologists add things to their reports to make themselves feel better, but it doesn't help anyone. It also manages to clutter things up and makes one come across as wishy-washy.

In contradistinction, I think it is helpful to tell someone to correlate for something specific, particularly if the imaging findings are equivocal. Not sure if that gas is in the bowel wall or not? Correlate with lactate. Not sure if that fracture is old or not? Correlate for point tenderness.
 
  • Like
Reactions: 1 users
I agree. I shadowed a DR during the winter break and he told me during his residency (early 2000’s) he was told that AI was going to “take over” the radiology field, yet he’s been practing for 10 years plus and he hasn’t seen any difference within the field. He told me to not pay attention to any of that, if I did choose DR as a specialty in the future.

Are there any publications from that era saying so, or is it just word of mouth?
 
  • Like
Reactions: 1 users
Are there any publications from that era saying so, or is it just word of mouth?

Yeah, just word of mouth from him. He never specified if there were any publications.
 
AI will make radiologists much more efficient, which will decrease the number of radiologists needed for any given amount of work.

It’s coming. If you won’t be an attending for 10 years, it might be an issue.

Then again, it will be an issue for every other non-surgical specialty soon enough too. And eventually the surgeons.

tasks.png


This is an XKCD comic from 2014 (made by a really smart former NASA engineer if you’re not familiar). Classifying something as a bird used to be really hard for a computer to do, very recently.

You can now solve this and similar questions with a few lines of code. That’s why the analogies to Mammo CAD are silly - any comparison from before 2012 doesn’t count.

We underestimate how complicated the visual tasks we find easy are thanks to evolution, and overestimate how hard more technical questions are.

We’re not used to it, but it’s easier for a computer to learn to identify a PE than a bird.

So would I recommend radiology to an MS1? Maybe if you have an MS in CS or plan to do IR. Otherwise, caveat emptor.


I agree, In 15 years it will be a lot different.
 
Some of us will stay relevant, not all of us.

And there may not need to be as many of us.

Your argument works equally well for the safety of lying down on train tracks.

I appreciate you contributing your more pessimistic view of the future of Radiology - I feel like when reading SDN and Aunt Minnie, people are really fast to say “nothing to worry about here, move along” when AI comes up. But I can’t help feeling that even if the sky was falling, people would be afraid to scare away students by telling the truth. Saying “don’t worry about it” feels disingenious given I’m almost 10 years away from my first real paycheck, and a lot can happen in 10 years with the current pace of technology.

That said, I have a hunch that this might actually be the perfect time to board the radiology train:
There literally are still hospitals using paper charts. “medicine moves slowly” is an understatement.

I expect we will see some actual progress with AI starting to be implemented in narrow capacities and having a non-negligible impact on clinical practice over these next 10 years. Those of us going through residency during this time will have “grown up” with that technology, and probably will get more exposure and practice using those tools than those who are already out in private practice.

I would bet my money that 30 years from now we’ll be on the downward side of the curve, with AI displacing many radiologists at least with regards to image interpretation, and assuming radiologists don’t find another way to stay relevant (which I imagine they would, given how strongly the field seems to be embracing AI compared to other fields threatened by it)... But I think we will be on the ascending side of the curve for quite a while yet, with demand for radiologists continuing to increase, and with AI actually increasing profits in the shorter term.

Imagine if tomorrow somebody said they had produced an AI that we could feed a scan and say “interpret this CT for me”... How long would it take to be approved by the FDA, medicolegal obstacles overcome, and then purchased and adopted by practices? Maybe 5 years at best, more likely 10 years before it would start actually replacing jobs? And from what I understand as a clueless M2 we are still a long way away from that point.

Anything is possible, but I have a hard time envisioning a scenario where I can’t manage at least a 20-year career as an attending DR from this point.

Agree/disagree?

(Wow, that got long-winded and rambley...)
 
  • Like
Reactions: 6 users
I appreciate you contributing your more pessimistic view of the future of Radiology - I feel like when reading SDN and Aunt Minnie, people are really fast to say “nothing to worry about here, move along” when AI comes up. But I can’t help feeling that even if the sky was falling, people would be afraid to scare away students by telling the truth. Saying “don’t worry about it” feels disingenious given I’m almost 10 years away from my first real paycheck, and a lot can happen in 10 years with the current pace of technology.

That said, I have a hunch that this might actually be the perfect time to board the radiology train:
There literally are still hospitals using paper charts. “medicine moves slowly” is an understatement.

I expect we will see some actual progress with AI starting to be implemented in narrow capacities and having a non-negligible impact on clinical practice over these next 10 years. Those of us going through residency during this time will have “grown up” with that technology, and probably will get more exposure and practice using those tools than those who are already out in private practice.

I would bet my money that 30 years from now we’ll be on the downward side of the curve, with AI displacing many radiologists at least with regards to image interpretation, and assuming radiologists don’t find another way to stay relevant (which I imagine they would, given how strongly the field seems to be embracing AI compared to other fields threatened by it)... But I think we will be on the ascending side of the curve for quite a while yet, with demand for radiologists continuing to increase, and with AI actually increasing profits in the shorter term.

Imagine if tomorrow somebody said they had produced an AI that we could feed a scan and say “interpret this CT for me”... How long would it take to be approved by the FDA, medicolegal obstacles overcome, and then purchased and adopted by practices? Maybe 5 years at best, more likely 10 years before it would start actually replacing jobs? And from what I understand as a clueless M2 we are still a long way away from that point.

Anything is possible, but I have a hard time envisioning a scenario where I can’t manage at least a 20-year career as an attending DR from this point.

Agree/disagree?

(Wow, that got long-winded and rambley...)

You are a mature MS2. You will do well.
 
  • Like
Reactions: 2 users
I appreciate you contributing your more pessimistic view of the future of Radiology - I feel like when reading SDN and Aunt Minnie, people are really fast to say “nothing to worry about here, move along” when AI comes up. But I can’t help feeling that even if the sky was falling, people would be afraid to scare away students by telling the truth. Saying “don’t worry about it” feels disingenious given I’m almost 10 years away from my first real paycheck, and a lot can happen in 10 years with the current pace of technology.

That said, I have a hunch that this might actually be the perfect time to board the radiology train:
There literally are still hospitals using paper charts. “medicine moves slowly” is an understatement.

I expect we will see some actual progress with AI starting to be implemented in narrow capacities and having a non-negligible impact on clinical practice over these next 10 years. Those of us going through residency during this time will have “grown up” with that technology, and probably will get more exposure and practice using those tools than those who are already out in private practice.

I would bet my money that 30 years from now we’ll be on the downward side of the curve, with AI displacing many radiologists at least with regards to image interpretation, and assuming radiologists don’t find another way to stay relevant (which I imagine they would, given how strongly the field seems to be embracing AI compared to other fields threatened by it)... But I think we will be on the ascending side of the curve for quite a while yet, with demand for radiologists continuing to increase, and with AI actually increasing profits in the shorter term.

Imagine if tomorrow somebody said they had produced an AI that we could feed a scan and say “interpret this CT for me”... How long would it take to be approved by the FDA, medicolegal obstacles overcome, and then purchased and adopted by practices? Maybe 5 years at best, more likely 10 years before it would start actually replacing jobs? And from what I understand as a clueless M2 we are still a long way away from that point.

Anything is possible, but I have a hard time envisioning a scenario where I can’t manage at least a 20-year career as an attending DR from this point.

Agree/disagree?

(Wow, that got long-winded and rambley...)

In the 1950s, systematic chemo was on the rise, and after the inital cure of certain hematological cancers, people declared radiation oncology to be obsolete....

70 years later, radonc is still alive and kicking.

AI can well take over my job in 15 years, but by then I’ll be an IR hopefully. Or maybe it won’t in 50 years.
 
  • Like
Reactions: 1 user
So we’re quoting irrelevant things now?

How about the industrial revolution? John Henry? The cotton gin? The printing press?

Was your car hand made? How about your clothes?

In the medical field, I’m sure the leech breeders ended up having a hard time.

Let’s not pretend that technology can’t shift displacing entire classes of workers. Industrial robots got the blue collars, better AI just might get the white collars.

Are you a radiology attending? Because I am directly reporting the opinion of one that the AI threat is overrated.

Can you prove literature to support impending ability of AI to intrepret images? Not the chest net one plz. The one has been debunked.
 
Are you a radiology attending? Because I am directly reporting the opinion of one that the AI threat is overrated.

Can you prove literature to support impending ability of AI to intrepret images? Not the chest net one plz. The one has been debunked.

In his recent video lecture on AI in Radiology, Elliot Fishman (attending from Hopkins) cites about 20 studies suggesting that the AI takeover is "inevitable" and rapidly approaching. With regards to the question, "Will we need more or less Radiologists in a decade?" he hedges by saying "we'll have to see, I don't want to guess that."

He ends part one of the video on an optimistic note:
"You can say, like anything, when you have change there will be winners and losers, but I think if you go along with AI and you work and you change, then I think you're gonna be one of the winners".

 
  • Like
Reactions: 1 user
How are you directly reporting someone else’s opinion? And what does being an attending have to do with anything (not saying I’m not one)?



How has ChestNet been debunked again? There are huge problems with the dataset labels, but I doubt that’s what you’re referring to...

Recommend you to actually do some diagnostic radiology before commenting on AI. I am sitting right next to a family member who is an radiology attending and is of the opinion that AI will affect our field but in time frame of 50 years, not 10.
 
  • Like
Reactions: 1 user
:lol:

Sips coffee. Reads another case. Laughs at med student. Counts doubloons.

Doubt a radiology attending would have such a lack of understanding over limitation of backpropagation. Or perhaps you think AI can easily interpret a post ops spine MRI too :))
 
Most radiology attendings don’t know what backpropagation is. Do you?

s25RsOr%2B-%2BImgur.gif

Ds_DVEvWkAAN3tn.jpg


And yes, I definitely think most spine imaging will eventually be automated.


How do you have so much time on your hands to be on SDN if you're an attending? Seems odd...
 
Cool stuff. Implementation of this in clinical practice is 40-50 years away.

Could you explain where you draw that timeline of "40-50 years" from? This technology literally has already been implemented in clinical practice - and tangibly improved patient outcomes - in a prospective trial that took place almost 1 year ago. As far as I can tell, cost and logistics are the only reason AI isn't already being used for improving workflow. The AI itself is already there for some use cases:

Advanced machine learning in action: identification of intracranial hemorrhage on computed tomography scans of the head with clinical workflow integration

"The predictive model was implemented prospectively for 3 months to re-prioritize “routine” head CT studies as “stat” on realtime radiology worklists if an ICH was detected"

"During implementation, 94 of 347 “routine” studies were re-prioritized to “stat”, and 60/94 had ICH identified by the radiologist. Five new cases of ICH were identified, and median time to diagnosis was significantly reduced (p < 0.0001) from 512 to 19 min. In particular, one outpatient with vague symptoms on anti-coagulation was found to have an ICH which was treated promptly with reversal of anticoagulation, resulting in a good clinical outcome."

"In conclusion, an artificial intelligence algorithm can prioritize radiology worklists to reduce time to diagnosis of new outpatient ICH by 96% and may also identify subtle ICH overlooked by radiologists. This demonstrates the positive impact of advanced machine learning in radiology workflow optimization."
 
  • Like
Reactions: 1 user
Could you explain where you draw that timeline of "40-50 years" from? This technology literally has already been implemented in clinical practice - and tangibly improved patient outcomes - in a prospective trial that took place almost 1 year ago. As far as I can tell, cost and logistics are the only reason AI isn't already being used for improving workflow. The AI itself is already there for some use cases:

Advanced machine learning in action: identification of intracranial hemorrhage on computed tomography scans of the head with clinical workflow integration

AI has been talked about for YEARS. In pathology too. All these studies are cool but actual implementation into clinical practice? Nah. That'll take forever. If you take into consideration how slow medicine is to market changes, the cost, the legal hurdle, I don't see it happening anytime soon in any big capacity. Maybe small applications like measuring nodules, which will increase efficiency and therefore pay for my generation. It won't really affect me in my lifetime. Ultimately, radiologists have always found a way to stay relevant so not too worried.
 
  • Like
Reactions: 2 users
AI has been talked about for YEARS. In pathology too. All these studies are cool but actual implementation into clinical practice? Nah. That'll take forever. If you take into consideration how slow medicine is to market changes, the cost, the legal hurdle, I don't see it happening anytime soon in any big capacity. Maybe small applications like measuring nodules, which will increase efficiency and therefore pay for my generation. It won't really affect me in my lifetime. Ultimately, radiologists have always found a way to stay relevant so not too worried.


Implementation already started.
Lots of start ups out there
Check out AIDOC. They have a software to check for hemorrhage on head CT
 
  • Like
Reactions: 1 user
Could you explain where you draw that timeline of "40-50 years" from? This technology literally has already been implemented in clinical practice - and tangibly improved patient outcomes - in a prospective trial that took place almost 1 year ago. As far as I can tell, cost and logistics are the only reason AI isn't already being used for improving workflow. The AI itself is already there for some use cases:

Advanced machine learning in action: identification of intracranial hemorrhage on computed tomography scans of the head with clinical workflow integration
The senior author on this paper is VP healthcare at Google and former bigwig at Geisinger.

They are further along than I expected.

I expect AI will begin affecting academic centers in 5-10 years. We already have an AI for bone age. The question is when something FDA approved and productized by Primordial/Nuance for sale into existing installations for private practice.
 
  • Like
Reactions: 1 user
Excellent post.

I’m a resident in a different field (neurology) and I was very close to choosing radiology for the exact reasons you mentioned above. I ended up pursuing neurology because patient interaction was very important to me.
 
It’ll become a question of who buys and controls the software, at least initially. I imagine, and hope, it will be on hospitals/systems to purchase and implement with contracted radiologists overseeing it, just how we oversee hospital owned equipment and hospital employed technologists currently.

Where it could get...interesting...for private practice is if corporations like Envision purchase software packages and make them included as part of their takeover offer. I imagine the price tag on this stuff is going to be quite high. And this sort of development would increase pressure to corporatize.
 
  • Like
Reactions: 1 user
The exception is probably breast imaging, where someone willing to do that can become partners and/or make more money while still being part-time.
I hear this often about breast imaging. Why is that? Is there some reason that it is a very undesirable fellowship/specialty so no one wants to go into it? This just a supply/demand thing?
 
I hear this often about breast imaging. Why is that? Is there some reason that it is a very undesirable fellowship/specialty so no one wants to go into it? This just a supply/demand thing?

Maybe the subject matter (breasts, occasionally lymph nodes), yield (the nature of screening), or practice style (talking to patients)?
 
Maybe the subject matter (breasts, occasionally lymph nodes), yield (the nature of screening), or practice style (talking to patients)?

Once you read a mammo screener you would understand
 
I hear this often about breast imaging. Why is that? Is there some reason that it is a very undesirable fellowship/specialty so no one wants to go into it? This just a supply/demand thing?

Maybe the subject matter (breasts, occasionally lymph nodes), yield (the nature of screening), or practice style (talking to patients)?

Yeah, Cognovi hit the major points. It's a subspecialty that cares almost exclusively about a single disease of a single organ that, with rare (1%) exception, exists in only half of the population to begin with. That disease is highly publicized and highly emotionalized. And the average person, even the average well-educated person, has a poor understanding of the nature and limitations of the imaging. It adds up to make for a litigious environment.

Of course, some people like the narrow breadth, but most don't. Other good things is that it's a decent mix of modalities and, often times, procedures. It's an opportunity to interact more with patients, if that's your thing, but for most of us it's not, especially regarding such an emotional topic. Lastly, it can give the radiologist greater control over the patient, because it really is your responsibility to work up the patient appropriately, as codified by law.
 
  • Like
Reactions: 1 users
Yeah, Cognovi hit the major points. It's a subspecialty that cares almost exclusively about a single disease of a single organ that, with rare (1%) exception, exists in only half of the population to begin with. That disease is highly publicized and highly emotionalized. And the average person, even the average well-educated person, has a poor understanding of the nature and limitations of the imaging. It adds up to make for a litigious environment.

Of course, some people like the narrow breadth, but most don't. Other good things is that it's a decent mix of modalities and, often times, procedures. It's an opportunity to interact more with patients, if that's your thing, but for most of us it's not, especially regarding such an emotional topic. Lastly, it can give the radiologist greater control over the patient, because it really is your responsibility to work up the patient appropriately, as codified by law.
So, if you are interested in radiology, but still want the opportunity to occasionally do an actual "workup" of a patient (listen to their heart and lungs, talk to them, etc.) would mammography be a solution?
 
So, if you are interested in radiology, but still want the opportunity to occasionally do an actual "workup" of a patient (listen to their heart and lungs, talk to them, etc.) would mammography be a solution?

The only reasons you as a radiologist are going to listen to a patient's chest
  • you're planning to or just did put a needle in their chest
  • you're planning to sedate them
  • you're evaluating someone who is actively having an allergic-like reaction to contrast media.
 
  • Like
Reactions: 1 users
I highly recommend radiology. It has its pros and cons as with any field. It’s broad enough that you can find a subspecialty you like. The jobs are very diverse. You can spend most of your time in one subspecialty or practice the full breadth of radiology.
 
  • Like
Reactions: 1 user
I'm on mammo as a first year resident, and so far it is great. To the poster above who commented on, what I assume to be the tediousness of reviewing screeners, isn't reviewing any normal exam, tedious?
 
  • Like
Reactions: 1 user
The only reasons you as a radiologist are going to listen to a patient's chest
  • you're planning to or just did put a needle in their chest
  • you're planning to sedate them
  • you're evaluating someone who is actively having an allergic-like reaction to contrast media.
Cool so your stethoscope doesn't collect dust if you're doing procedures? Good to know
 
Top