Dermatology or Diagnostic Radiology

[MillerDCorgi]

Hey All,

I'm halfway through my third year and really torn between two specialties: dermatology and radiology. I've had limited exposure to each field (just a few weeks), but I find both interesting and think I would be happy in either. However, since both fields are competitive, I really need to make a decision soon so I can start lining up appropriate letters of recommendation, away rotations, etc.

I am hoping to get some advice from those who have some experience in these fields. Below are some pros and cons of each. Please let me know if my perceptions are accurate or if I'm making false assumptions.

Dermatology Pros
- I enjoy the diagnostic process involved in identifying skin lesions
- I find dermatopathology interesting
- Love the challenge of diagnosing a cutaneous manifestation of a systemic disease
- Get satisfaction from helping those with common skin conditions (acne, eczema, etc.)
- Great lifestyle

Dermatology Cons
- I have pretty bad hands and don't enjoy procedures
- Although I enjoy patient care, it may be exhausting seeing 40+ patients per day for years on end
- In my limited experience, most practices are not terribly diverse; treating only a small number of skin conditions over and over again might become a grind

Diagnostic Radiology Pros
- I'm an introvert, and I would prefer seeing 5 patients a day to 40 patients per day
- Great deal of diversity between reading x-rays, CT, MRI, etc. with a wide range of pathologies
- Enjoy 'doc-to-doc' consults more than providing patient education
- Evolving field with new imaging techniques in development
- Teleradiology is appealing from a lifestyle perspective

Diagnostic Radiology Cons
- I have atrocious visuospatial intelligence (almost failed anatomy in the pre-clinical years)
- 'The Great Unknown' - radiology is so different from any other fields, I feel like I don't really know what I'd be getting into
- Job market is questionable
- Significantly longer training than dermatology (as I understand a fellowship is virtually required these days)
- Outsourcing?
- Concerns about losing the clinical skills developed in medical school (and intern year)

I imagine a lot of what I have written above is inaccurate; we only get a few weeks of electives in MS3, so my perceptions are based on very limited experience in each field.

Thoughts, suggestions, corrections, etc. would be greatly appreciated! Thank you very much for taking the time to read this.

---

Members don't see this ad.

 

[sovereign0]

Dermatology Cons
- I have pretty bad hands and don't enjoy procedures

I got some bad news for you - Diagnostic Rads does a significant amount of procedures too.

Also, while both specialties are "competitive" - Dermatology is worlds apart from DR. Not only do you need the Step score, research and aways are going to play a factor too.

My understanding for DR is that if you have the Step score, you'll match. For Derm, you can have the step score and still only have a 50%-65% shot of matching.

 

[Redpancreas]

Hey All,

I'm halfway through my third year and really torn between two specialties: dermatology and radiology. I've had limited exposure to each field (just a few weeks), but I find both interesting and think I would be happy in either. However, since both fields are competitive, I really need to make a decision soon so I can start lining up appropriate letters of recommendation, away rotations, etc.

I am hoping to get some advice from those who have some experience in these fields. Below are some pros and cons of each. Please let me know if my perceptions are accurate or if I'm making false assumptions.

Dermatology Pros
- I enjoy the diagnostic process involved in identifying skin lesions
- I find dermatopathology interesting
- Love the challenge of diagnosing a cutaneous manifestation of a systemic disease
- Get satisfaction from helping those with common skin conditions (acne, eczema, etc.)
- Great lifestyle

Dermatology Cons
- I have pretty bad hands and don't enjoy procedures
- Although I enjoy patient care, it may be exhausting seeing 40+ patients per day for years on end
- In my limited experience, most practices are not terribly diverse; treating only a small number of skin conditions over and over again might become a grind

Diagnostic Radiology Pros
- I'm an introvert, and I would prefer seeing 5 patients a day to 40 patients per day
- Great deal of diversity between reading x-rays, CT, MRI, etc. with a wide range of pathologies
- Enjoy 'doc-to-doc' consults more than providing patient education
- Evolving field with new imaging techniques in development
- Teleradiology is appealing from a lifestyle perspective

Diagnostic Radiology Cons
- I have atrocious visuospatial intelligence (almost failed anatomy in the pre-clinical years)
- 'The Great Unknown' - radiology is so different from any other fields, I feel like I don't really know what I'd be getting into
- Job market is questionable
- Significantly longer training than dermatology (as I understand a fellowship is virtually required these days)
- Outsourcing?
- Concerns about losing the clinical skills developed in medical school (and intern year)

I imagine a lot of what I have written above is inaccurate; we only get a few weeks of electives in MS3, so my perceptions are based on very limited experience in each field.

Thoughts, suggestions, corrections, etc. would be greatly appreciated! Thank you very much for taking the time to read this.

Honestly, it doesn't sound like you'd like either. You'd definitely have an easier time matching Radiology though.

 

[CherryRedDracul]

DR is going to have plenty of procedures in residency that you will need to do, but if you're out in private practice, you may be able to punt everything to the IR-trained guy.

 

[PatsyStone]

Dermatology Cons

- Although I enjoy patient care, it may be exhausting seeing 40+ patients per day for years on end


Diagnostic Radiology Pros

- I'm an introvert, and I would prefer seeing 5 patients a day to 40 patients per day

These are not insignificant points

 

[Pathman1000]

have you considered path? Sure it has its issues, but it sounds like what you may be looking for.

 

[DrfluffyMD]

Honestly, I would consider heme-onc or medicine subspecialty. Don’t jump on the rad train because jobs are easy to come by now. I bought the dip in radiology when job market was in dumpsters. Who knows how it will be like 6+ years from now.

 

[CherryRedDracul]

have you considered path? Sure it has its issues, but it sounds like what you may be looking for.

Job market is still turrible for path as it has been for a long time.

 

[19pieces19]

Honestly, I would consider heme-onc or medicine subspecialty. Don’t jump on the rad train because jobs are easy to come by now. I bought the dip in radiology when job market was in dumpsters. Who knows how it will be like 6+ years from now.

Do you really recommend current students not to pursue your field?

 

[libertyyne]

If you are competitive enough for derm. Go derm. Don't worry about the good with your hands part as residency will resolve that piece.

 

[MillerDCorgi]

Thank you all for the insightful responses! I appreciate the time. Responses follow:

I got some bad news for you - Diagnostic Rads does a significant amount of procedures too.

Also, while both specialties are "competitive" - Dermatology is worlds apart from DR. Not only do you need the Step score, research and aways are going to play a factor too.

My understanding for DR is that if you have the Step score, you'll match. For Derm, you can have the step score and still only have a 50%-65% shot of matching.

This is exactly why I asked this question: at the practice in which I did my brief DR rotation, the IR-trained radiologists performed most of the procedures. The DR-only guys did a couple (stereotactic breast biopsies, some U/S-guided biopsies), but it was pretty limited. It sounds like that is not the norm?

Honestly, it doesn't sound like you'd like either. You'd definitely have an easier time matching Radiology though.

It's really hard to know which field is a good choice with some little exposure.

I am fairly confident that I would be able to match dermatology based on Step score, research, and connections. At the moment, I'm more concerned about figuring out which field I really like!

DR is going to have plenty of procedures in residency that you will need to do, but if you're out in private practice, you may be able to punt everything to the IR-trained guy.

Thanks. That was my experience based on a very short rotation.

These are not insignificant points

Agreed. My personality type is probably better suited for DR. However, I do enjoy some patient interaction, and find the idea of splitting time between clinical dermatology and dermatopathology appealing. Is 40 patients pretty typical for dermatology? I've seen a wide range at different practices.

have you considered path? Sure it has its issues, but it sounds like what you may be looking for.

I absolutely did! Unfortunately, 5 out of the 6 pathologists I worked with on my rotation discouraged me from applying due to the job market. Kind of took the wind out of my sails.

Honestly, I would consider heme-onc or medicine subspecialty. Don’t jump on the rad train because jobs are easy to come by now. I bought the dip in radiology when job market was in dumpsters. Who knows how it will be like 6+ years from now.

Interestingly, hem/onc was one medicine subspecialty I was considering. I did not have the chance to do an elective in the field, but haven't totally crossed it off my list.

Other than the uncertain job market, why would you suggest hem/onc or a medicine subspecialty over DR?

Job market is still turrible for path as it has been for a long time.

That's what I was told as well.

If you are competitive enough for derm. Go derm. Don't worry about the good with your hands part as residency will resolve that piece.

Thanks. I imagine my hands would get better over time. I might even like procedures at some point. Procedures aside, why do you suggest dermatology over DR?

 

[libertyyne]

Thank you all for the insightful responses! I appreciate the time. Responses follow:



This is exactly why I asked this question: at the practice in which I did my brief DR rotation, the IR-trained radiologists performed most of the procedures. The DR-only guys did a couple (stereotactic breast biopsies, some U/S-guided biopsies), but it was pretty limited. It sounds like that is not the norm?



It's really hard to know which field is a good choice with some little exposure.

I am fairly confident that I would be able to match dermatology based on Step score, research, and connections. At the moment, I'm more concerned about figuring out which field I really like!



Thanks. That was my experience based on a very short rotation.



Agreed. My personality type is probably better suited for DR. However, I do enjoy some patient interaction, and find the idea of splitting time between clinical dermatology and dermatopathology appealing. Is 40 patients pretty typical for dermatology? I've seen a wide range at different practices.



I absolutely did! Unfortunately, 5 out of the 6 pathologists I worked with on my rotation discouraged me from applying due to the job market. Kind of took the wind out of my sails.



Interestingly, hem/onc was one medicine subspecialty I was considering. I did not have the chance to do an elective in the field, but haven't totally crossed it off my list.

Other than the uncertain job market, why would you suggest hem/onc or a medicine subspecialty over DR?



That's what I was told as well.



Thanks. I imagine my hands would get better over time. I might even like procedures at some point. Procedures aside, why do you suggest dermatology over DR?

Less training. Good lifestyle. Low stress. Better job market. In order to get a radiology job I hear fellowship is kind of the norm nowm.And low risk of AI vs rads .

 

[Pathman1000]

Job market is still turrible for path as it has been for a long time.

It's terrible for terrible candidates.

Less training. Good lifestyle. Low stress. Better job market. In order to get a radiology job I hear fellowship is kind of the norm nowm.And low risk of AI vs rads .

The job market is always an issue but it's not as bad as it is made out to be, especially for an English-speaking AMG from a reputable program. The people I see who struggle the most are generally those I wouldn't personally want to hire either. It's probably not all that much worse than radiology or rad-onc in reality. I wouldn't let the job market alone be the sole deciding factor in what you will be doing for the next 3-4 decades.

 

[Lnsean]

It's terrible for terrible candidates.


The job market is always an issue but it's not as bad as it is made out to be, especially for an English-speaking AMG from a reputable program. The people I see who struggle the most are generally those I wouldn't personally want to hire either. It's probably not all that much worse than radiology or rad-onc in reality. I wouldn't let the job market alone be the sole deciding factor in what you will be doing for the next 3-4 decades.

I mean it's a very valid worry. Who wants to spend 6 years of specialized training to have the job market be in shtter...and it's not like he's in love with radiology. OP, go derm if you're competitive for it. Only do rads if you can't see yourself doing anything else.

 

[seapbu10]

If you're a top candidate for rads and match at a top 20 place, jobs will not be a problem for you in any job market. Can you get 10-18 wks vacay/yr in derm with top specialist pay (median salary >500)?

AI will never replace radiologists and will only help with efficiency. CAD anyone (mammo)? You can't hold a computer program liable for malpractice...

Radiology is so much more interesting. You get to see everything you've read about in med school and a heck of a lot more. You get to see all of the cool cases that come into the hospital.. you get to problem solve and help decipher medical mysteries.. Rads is really awesome and it's a shame how little we're exposed to as med students and it's unfortunate how limited/boring the experience on rotation is for students

 

[libertyyne]

If you're a top candidate for rads and match at a top 20 place, jobs will not be a problem for you in any job market. Can you get 10-18 wks vacay/yr in derm with top specialist pay (median salary >500)?

AI will never replace radiologists and will only help with efficiency. CAD anyone (mammo)? You can't hold a computer program liable for malpractice...

Every person who has been replaced by computers says the same thing.
https://arxiv.org/pdf/1711.05225.pdf

 

[seapbu10]

Exactly an example for AI making the radiologist's life a little easier.. (maybe) improving the possible pneumonia vs atelectasis determination which can never be 100% correct, however, there's a lot more on a cxr than that.. plus there are a ton of different modalities with many more in the pipeline.. AI with CAD for mammo is currently abysmal (less than 50% accuracy) and it is highly unlikely to replace any mammographer considering how much is at stake/liability involved.. definitely needs oversight.. AI is highly unlikely to take over radiologists' jobs during our careers

 

[libertyyne]

Exactly an example for AI making the radiologist's life a little easier.. (maybe) improving the possible pneumonia vs atelectasis determination which can never be 100% correct, however, there's a lot more on a cxr than that.. plus there are a ton of different modalities with many more in the pipeline.. AI with CAD for mammo is currently abysmal (less than 50% accuracy) and it is highly unlikely to replace any mammographer considering how much is at stake/liability involved.. definitely needs oversight.. AI is highly unlikely to take over radiologists' jobs during our careers

I have the opposite perception. AI is getting better, people like to point towards ekg and old tech that was designed by coders and say see it cant work. When the reality is it is an eventuality that it will become better than humans are, and be cheaper. If it can 25% of scans it makes it that radiologists read less and earn less or you need less radiologists to read more scans. This does not end in any way that is good for radiologists.

 

[seapbu10]

I have the opposite perception. AI is getting better, people like to point towards ekg and old tech that was designed by coders and say see it cant work. When the reality is it is an eventuality that it will become better than humans are, and be cheaper. If it can 25% of scans it makes it that radiologists read less and earn less or you need less radiologists to read more scans. This does not end in any way that is good for radiologists.

A radiologist does more than just pick out findings. They put things into clinical context and determine whether a finding is even important or not. EKGs are always interpreted by a physician to put things into clinical context. Imaging volume has expanded exponentially in number year by year, especially with the ever increasing number of graduating NPs who often order extra unnecessary studies.. the expansion of new MRI sequences and images/study has also increased.. on the other hand when talking about job replacement, I see more and more NPs practicing dermatology and gaining more independence

 

[libertyyne]

A radiologist does more than just pick out findings. They put things into clinical context and determine whether a finding is even important or not. EKGs are always interpreted by a physician to put things into clinical context. Imaging volume has expanded exponentially in number year by year, especially with the ever increasing number of graduating NPs who often order extra unnecessary studies.. the expansion of new MRI sequences and images/study has also increased.. on the other hand when talking about job replacement, I see more and more NPs practicing dermatology and gaining more independence

Why take a career gamble starting out? 30 years is a long time , especially for technology to change.

 

[seapbu10]

A radiologist does more than just pick out findings. They put things into clinical context and determine whether a finding is even important or not. EKGs are always interpreted by a physician to put things into clinical context. Imaging volume has expanded exponentially in number year by year, especially with the ever increasing number of graduating NPs who often order extra unnecessary studies.. the expansion of new MRI sequences and images/study has also increased.. on the other hand when talking about job replacement, I see more and more NPs practicing dermatology and gaining more independence

Why take a career gamble starting out? 30 years is a long time , especially for technology to change.

It's really not gambling at all if you choose something you really enjoy.. and if you're financially smart, you won't have to work for 30 years and can be financially independent well before then enabling you to pursue other life interests, especially as a private practice radiologist

 

[seapbu10]

It's really not gambling at all if you choose something you really enjoy.. and if you're financially smart, you won't have to work for 30 years and can be financially independent well before then enabling you to pursue other life interests, especially as a private practice radiologist

Good luck, op! Both are great fields in their own way!

 

[Decicco]

Diagnostic radiology will be disrupted by AI and machine learning.

 

[GadRads]

Every person who has been replaced by computers says the same thing.
https://arxiv.org/pdf/1711.05225.pdf

I quickly read the article and the performance of the model is quite impressive, but it misses the mark in some significant ways. As a radiology resident, I have come to know that physicians and medical students without significant exposure to radiology really overestimate their knowledge and expertise when it comes to imaging. Many don't understand what exactly a radiologist does as he/she flips through images. For instance, most chest X-rays I read are suboptimal and can be hard to read, and not of the quality used in that article, which are are quite easy.

The article also conflates the appearance of different pathologies as seen in Table 2. This is why the article misses the point. It puts a list of known diagnoses and assumes a chest x-ray can frequently allow confident diagnosis? Table 2 confirms to me that the authors have a poor understanding of pathology. For example, they treat pneumonia or consolidation as separate entities, which is dubious. Second, what if the patient has an obstructive pneumonia with additional lobar segmental atelectasis? You think a chest x-ray can confidently distinguish either, or a thymoma from a lymphoma without additional imaging or work-up? The article also fails to clarify that pneumonia is a clinical diagnosis. Medicine isn't so binary.

I expected radiology residency to be quite difficult, but it has been even more challenging, more so than my internal medicine intern year. I have little worry about AI and I expect AI to help us do our jobs better and faster.

 

[GadRads]

A radiologist does more than just pick out findings. They put things into clinical context and determine whether a finding is even important or not. EKGs are always interpreted by a physician to put things into clinical context. Imaging volume has expanded exponentially in number year by year, especially with the ever increasing number of graduating NPs who often order extra unnecessary studies.. the expansion of new MRI sequences and images/study has also increased.. on the other hand when talking about job replacement, I see more and more NPs practicing dermatology and gaining more independence

In my hospital, NP's staff oncology, cardiology, IM and dermatology. I see them write notes, do H&P's, place orders, and so on, following by the attending's "I saw and examined the patient and agree with the plan". I have yet to see them interpret imaging, but they misuse imaging, particularly the outpatient midlevel providers. Only ultrasound techs come close since they need to know what to scan, and that is only in ultrasound. Nuc med techs come in second. However none can integrate the data with CT, MRI, etc and the overall clinical picture.

Radiology really requires strong medical knowledge (including a lot of the tiny details from MS1 and MS2), as well as understand the clinical picture. You need to memorize a lot of information and critically think about it. This is why subspecialty training is essential as it is impossible to know all of it.

I feel you can BS your way in internal medicine, hard to do that in radiology.

 

[radsisrad]

If you're a top candidate for rads and match at a top 20 place, jobs will not be a problem for you in any job market. Can you get 10-18 wks vacay/yr in derm with top specialist pay (median salary >500)?

Just so this is clear to med students: the reason why Rads tends to come with (relatively) so much vacation time is that when you're working, your brain is engaged nonstop. That extra break time appears to be crucial, based on my discussions with Rads residents and attendings.

 

[Apoplexy__]

Just so this is clear to med students: the reason why Rads tends to come with (relatively) so much vacation time is that when you're working, your brain is engaged nonstop. That extra break time appears to be crucial, based on my discussions with Rads residents and attendings.

To be fair, you have to have a very high IQ to be a radiologist and to require that much vacation.

 

[username456789]

.

 

[MedicalDoge]

What is wrong with the pathology job market?

 

[Cognovi]

Lol what do you think other doctors are doing? Shutting off their brains with certain patients and turning it on for others?

Charting.

 

[libertyyne]

What is wrong with the pathology job market?

oversupply of pathologists, pathology assistants, automation, long careers for current pathologists, saturation in urban and desirable areas.

 

[Brorthopedic]

What is wrong with the pathology job market?

 

[MedicalDoge]

Hey haven't even started med school yet sorry I'm not looking at specialty job markets lol.

The pathologist I shadowed said that pathology would be a great field for me to go in to, I guess he's been out the game for a while..

 

[CherryRedDracul]

My question is so what? With the rate at which medical imaging technology advances, it seems to me there will still be need for diagnostic radiologists at the cutting edge of the field, supervising departments, verifying/training machine learning (you can't just set and forget a machine learning algorithm, it takes a lot to train one, and then it must adapt), and making judgement calls when AI isn't able to make a confident decision. The job market will be smaller, but the job won't be any less interesting, and might even be better to have a break from the dark room.

I expect diagnostic radiology to evolve with technology. I'm not worried about whether I'll be replaced by machines in the near future.

 

[libertyyne]

My question is so what? With the rate at which medical imaging technology advances, it seems to me there will still be need for diagnostic radiologists at the cutting edge of the field, supervising departments, verifying/training machine learning (you can't just set and forget a machine learning algorithm, it takes a lot to train one, and then it must adapt), and making judgement calls when AI isn't able to make a confident decision. The job market will be smaller, but the job won't be any less interesting, and might even be better to have a break from the dark room.

The whole point is that less radiologists will be needed to oversee that, this equates to less pay, less autonomy, less vacay time, less preferable locations etc based on current supply of radiologists. If I replace a truck driver with AI do i need the same number of truck drivers around?

 

[DrfluffyMD]

The whole point is that less radiologists will be needed to oversee that, this equates to less pay, less autonomy, less vacay time, less preferable locations etc based on current supply of radiologists. If I replace a truck driver with AI do i need the same number of truck drivers around?

Probably. Same reason why we need similar amount of pilots despite planes been flying themselves into space and back since 1989

 

[libertyyne]

Probably. Same reason why we need similar amount of pilots despite planes been flying themselves into space and back since 1989

once again pointing to hardcoded automation systems and their limitations.

 

[MillerDCorgi]

Thank you all for your thoughts! I appreciate everyone's insight.

I'm still very much on the fence between dermatology and DR, but leaning slightly toward the latter, as I think that my personality type is most compatible with DR. I do love the diagnostic aspect of dermatology. However, although it's traditionally considered a "lifestyle specialty," a 6- or 8-hour day that involves jumping from room to room and seeing dozens of patients would be draining.

I'm not terribly concerned about the effect of AI on DR. In fact, my understanding is that AI technology will be able to accurately identify skin lesions much more quickly than it will be able to analyze CT scans and MRIs.

Having said all of that, I am still apprehensive about committing to DR simply because the training process sounds inordinately challenging. I mentioned earlier that I have atrocious visuospatial intelligence. Is this something that can be developed over time, or would a career in DR condemn me to years of banging my head against the wall in hopelessness? I am confident that I could become a good dermatologist, in spite of my relatively poor dexterity, but I am uncertain if I possess the skills required to be successful in DR.

Thanks once again for all of your thoughts! I apologize for not responding to each post, but I have read every one.

 

[libertyyne]

Thank you all for your thoughts! I appreciate everyone's insight.

I'm still very much on the fence between dermatology and DR, but leaning slightly toward the latter, as I think that my personality type is most compatible with DR. I do love the diagnostic aspect of dermatology. However, although it's traditionally considered a "lifestyle specialty," a 6- or 8-hour day that involves jumping from room to room and seeing dozens of patients would be draining.

I'm not terribly concerned about the effect of AI on DR. In fact, my understanding is that AI technology will be able to accurately identify skin lesions much more quickly than it will be able to analyze CT scans and MRIs.

Having said all of that, I am still apprehensive about committing to DR simply because the training process sounds inordinately challenging. I mentioned earlier that I have atrocious visuospatial intelligence. Is this something that can be developed over time, or would a career in DR condemn me to years of banging my head against the wall in hopelessness? I am confident that I could become a good dermatologist, in spite of my relatively poor dexterity, but I am uncertain if I possess the skills required to be successful in DR.

Thanks once again for all of your thoughts! I apologize for not responding to each post, but I have read every one.

I haven't seen any studies regarding the attrition rate for rads residencies. But I have a suspicion that they are in line with other specialties. They don't test spatial reasoning prior to admittance to residency so I dont think you should be apprehensive about that. The skills will come with practice it might take you longer.
You are right about derm images, although there is an aspect of treatment/biopsy that won't be automated. Have you thought about IR?

 

[masaraksh]

omg radiology residents are like the single group that feels the need to bring up how challenging and mentally exhausting their field is. sure, it takes some smarts but from my rotations radiology residents arn't some form of gods gift to medicine or uber geniuses. If anything, the one unifying thing about rads residents is that they dont like working too much and care a lot about lifestyle. I don't care if I get flamed for this but its straight up true.

 

[19pieces19]

omg radiology residents are like the single group that feels the need to bring up how challenging and mentally exhausting their field is. sure, it takes some smarts but from my rotations radiology residents arn't some form of gods gift to medicine or uber geniuses. If anything, the one unifying thing about rads residents is that they dont like working too much and care a lot about lifestyle. I don't care if I get flamed for this but its straight up true.

I think that's moreso derm.

Rad still has to work 60 hour weeks.

 

[masaraksh]

Rad still has to work 60 hour weeks.

The horror.

But rads are the only ones who need to try to convince others how smart they are or how 'draining' the work is.

 

[CherryRedDracul]

omg radiology residents are like the single group that feels the need to bring up how challenging and mentally exhausting their field is. sure, it takes some smarts but from my rotations radiology residents arn't some form of gods gift to medicine or uber geniuses. If anything, the one unifying thing about rads residents is that they dont like working too much and care a lot about lifestyle. I don't care if I get flamed for this but its straight up true.

 

[LarryRad]

Every person who has been replaced by computers says the same thing.

This is either a joke or you have very poor knowledge on CXR interpretation. Ask any radiologist and they will tell you that the CXR is probabaly the most difficult examination to report!
Here are a few points:
- Pneumonia presents (generally) as airspace consolidation which has a wide differential including haemmorhage, aspiration and cancer e.g. lymphoma. It can be impossible to differentiate on a single CXR which is why we repeat imaging/advise further imaging. This means AI will have to produce a differential for airspace consolidation if it recognises it - this is long and sure to frustrate the referrer. Not to mention that a single CXR is likely to have simultaneous pathology e.g. infection on a background of heart failure or fibrosis (or both!).
- Pneumonia by definition requires the presence of NEW consolidation when compared to an old CXR. In this scenario AI would need to accurately compare two CXRs (if an older one is available) which may have different projections and temporal changes e.g. the patient may have had a mastectomy in the interim etc.
- I envisage much overcalling e.g. en face vessels being called nodules, vascular markings called fibrosis, skin folds called PTX.
- How will it deal with artefacts? Have you looked at ITU CXRs? (I've seen CXRs with IABP, CVC, Swanz-Ganz, valves, NGT, PPM/ICDs and ECG leads on the same radiograph!). Our hospital also offers ECMO.
That's just a little bit on the lungs! Then you have to consider the heart, main vessels, skeletal structures, soft tissues and infradiaphragmatic regions (I've diagnosed many fractures/dislocations, dissections, lymphadenopathy, PP and PM on CXRs).
- Even if AI learns to pick out abnormalities- it will probably give you a report full of lists of differentials for each abnormality. Hardly useful. You also need to know what to do with an abnormality - this requires medical knowledge (attending conferences etc.) which is constantly updating. Radiologists in the MDT guide our colleagues on what the next best step should be. In fact, I am interrupted at least 5-10 times a day by a physician or surgeon who would like my opinion. Somehow, I don't think speaking to a computer/AI will be as useful in these situations.

I can go on.

Trust me- AI will never replace the radiologist, it will be utilised but in a supportive manner. It's like the ECG machine giving everyone with Parkinson's atrial flutter due tot heir tremor- you will always need a human to review.

Whether you like it or not, radiologists are central to the diagnosis (and sometimes management) of many diseases. We are a useful resource and working together with our colleagues we can deliver optimal patient care.

 

[19pieces19]

This is either a joke or you have very poor knowledge on CXR interpretation. Ask any radiologist and they will tell you that the CXR is probabaly the most difficult examination to report!
Here are a few points:
- Pneumonia presents (generally) as airspace consolidation which has a wide differential including haemmorhage, aspiration and cancer e.g. lymphoma. It can be impossible to differentiate on a single CXR which is why we repeat imaging/advise further imaging. This means AI will have to produce a differential for airspace consolidation if it recognises it - this is long and sure to frustrate the referrer. Not to mention that a single CXR is likely to have simultaneous pathology e.g. infection on a background of heart failure or fibrosis (or both!).
- Pneumonia by definition requires the presence of NEW consolidation when compared to an old CXR. In this scenario AI would need to accurately compare two CXRs (if an older one is available) which may have different projections and temporal changes e.g. the patient may have had a mastectomy in the interim etc.
- I envisage much overcalling e.g. en face vessels being called nodules, vascular markings called fibrosis, skin folds called PTX.
- How will it deal with artefacts? Have you looked at ITU CXRs? (I've seen CXRs with IABP, CVC, Swanz-Ganz, valves, NGT, PPM/ICDs and ECG leads on the same radiograph!). Our hospital also offers ECMO.
That's just a little bit on the lungs! Then you have to consider the heart, main vessels, skeletal structures, soft tissues and infradiaphragmatic regions (I've diagnosed many fractures/dislocations, dissections, lymphadenopathy, PP and PM on CXRs).

I can go on.

Trust me- AI will never replace the radiologist, it will be utilised but in a supportive manner. It's like the ECG machine giving everyone with Parkinson's atrial flutter- you will always need a human to review.

Whether you like it or not, radiologists are central to the diagnosis (and sometimes management) of many diseases. We are a useful resource and working together with our colleagues we can deliver optimal patient care.

How many hours do you work per week?

 

[libertyyne]

This is either a joke or you have very poor knowledge on CXR interpretation. Ask any radiologist and they will tell you that the CXR is probabaly the most difficult examination to report!
Here are a few points:
- Pneumonia presents (generally) as airspace consolidation which has a wide differential including haemmorhage, aspiration and cancer e.g. lymphoma. It can be impossible to differentiate on a single CXR which is why we repeat imaging/advise further imaging. This means AI will have to produce a differential for airspace consolidation if it recognises it - this is long and sure to frustrate the referrer. Not to mention that a single CXR is likely to have simultaneous pathology e.g. infection on a background of heart failure or fibrosis (or both!).
- Pneumonia by definition requires the presence of NEW consolidation when compared to an old CXR. In this scenario AI would need to accurately compare two CXRs (if an older one is available) which may have different projections and temporal changes e.g. the patient may have had a mastectomy in the interim etc.
- I envisage much overcalling e.g. en face vessels being called nodules, vascular markings called fibrosis, skin folds called PTX.
- How will it deal with artefacts? Have you looked at ITU CXRs? (I've seen CXRs with IABP, CVC, Swanz-Ganz, valves, NGT, PPM/ICDs and ECG leads on the same radiograph!). Our hospital also offers ECMO.
That's just a little bit on the lungs! Then you have to consider the heart, main vessels, skeletal structures, soft tissues and infradiaphragmatic regions (I've diagnosed many fractures/dislocations, dissections, lymphadenopathy, PP and PM on CXRs).

I can go on.

Trust me- AI will never replace the radiologist, it will be utilised but in a supportive manner. It's like the ECG machine giving everyone with Parkinson's atrial flutter- you will always need a human to review.

Whether you like it or not, radiologists are central to the diagnosis (and sometimes management) of many diseases. We are a useful resource and working together with our colleagues we can deliver optimal patient care.

Even if we take the best case senario( for radiologists) where the software is only able to effectively make call's on negative X-rays for any pathology. Wouldnt that decrease the need for radiologists?

I am sure there are massive technological hurdles in implementation and regulatory hurdles. I still am not in the habit of betting against technology.

 

[LarryRad]

libertyyne, post: 19725000, member: 676376"]Even if we take the best case senario( for radiologists) where the software is only able to effectively make call's on negative X-rays for any pathology. Wouldnt that decrease the need for radiologists?

I am sure there are massive technological hurdles in implementation and regulatory hurdles. I still am not in the habit of betting against technology.[/QUOTE]

The issue here is a negative CXR can mean alot of things.
- I often overlook benign granulomas on a CXR. It is not an active lesion and nothing needs to be done. I won't report it and my colleagues won't worry about it. Will AI have the sense to do that- to know when to dismiss benign changes?
- A normal CXR for a young patient is different to a normal CXR for an elderly patient (for example it is common for the aorta to become tortuous with ageing)- AI would need to know the difference and assuming it has access to all the CXRs for a particular patient appreciate the normal change that occurs with ageing. Not to mention post-operative changes e.g. post lobectomy. Will it call rib resection from thoracotomy new rib fractures?!
- I find it very difficult to believe that any company would claim that their software is 100% perfect at ruling out all pathology without lawyering up to the max. Believe me some pathology can be very difficult to pick up on CXR like early fibrosis. There are other factors like composite shadowing and artefacts like companion shadows, which can confuse the picture. Forgive me, but I feel you are underestimating how difficult interpretation of a simple CXR can be!

 

[LarryRad]

How many hours do you work per week?

I work in the UK - as what you would call a senior resident?

 

[LarryRad]

Reduce the need for radiologists? I haven't even started talking about the other modalities-CT, USS, MR, fluoroscopy, Nuclear med, intervention etc. I report plain radiographs if I have spare time! You should visit your local radiology department.

 

[libertyyne]

Reduce the need for radiologists? I haven't even started talking about the other modalities-CT, USS, MR, fluoroscopy, Nuclear med, intervention etc. I report plain radiographs if I have spare time! You should visit your local radiology department.

i did, I was interested in radiology at one time.

I may be discounting the complexity involved in reading a scan, however you might be discounting the technological capabilities of AI.
Do you think AI can drive a car, surely there are limitless possibilties of nuance that need to be interpreted while doing that.
How about trade stocks? There are complexities involved in that as well.

Why do you think all imaging modalities are immune. I linked a paper earlier in the discussion that provided better reads compared to a radiologist for the same images for a number of conditions. Sure it isnt perfect but I am surprised that it doesnt give anyone pause.