5 ms·

It's an oversimplification to say AI will replace xyz job. It seems more likely that AI will simply sift through the data more thuroughly and look holistically

It's an oversimplification to say AI will replace xyz job. It seems more likely that AI will simply sift through the data more thuroughly and look holistically to catch things a radiologist might miss. A radiologist, for example, might miss spotting a small tumor in an xray taking for an unrelated hip injury. AI has a lot of complimentary value.

pharmakom 5y ago
I don’t see radiology work decreasing either. Instead, I think it will serve more people but at lower cost per person. No one will skip medical services if they can afford them, but currently prices are high. Imagine a future where a radiologist serves 10x customers as before by leveraging smart technologies, for similar overall compensation.
carbocation 5y ago
I agree with you and with the sibling comment!
cloverich 5y ago
It is a good point but I think one part is a bit backward in an ironic way. One reason AI hasn't replaced radiologists is because radiologists are typically very good physicians, and specifically do not look at images in isolation but review the record, talk with the physicians, sometimes the patients too, etc. So its actually backwards (in some cases) -- AI struggles because it looks at the image in isolation, while the radiologist is looking at the patient more holistically.
- koheripbal 5y ago
  Radiologists don't talk to patients (usually), so there's no reason why AI cannot be given all the same patient data. ...although reading doctors' notes is probably a whole 'nother AI program.
  - ska 5y ago
    > .although reading doctors' notes is probably a whole 'nother AI program. yep. One that also has a long history, and a lot of current players - an nobody has really good traction there either.
dx034 5y ago
But aren't we already over-diagnosing some cancers? Spotting more tiny tumors in unrelated images might do more harm (through procedures/treatment) than ignoring them. I'm not sure if we're really better off detecting every anomaly in someone's body.
- vecter 5y ago
  Why would we not want to know about a tumor in the body? I assume competent doctors will assess the risk of such a thing, but knowing about it is better than not.
  - vharuck 5y ago
    Doctors will optimize for patient outcomes, usually by doing all they can. Sometimes, this doesn't scale well. For example, the US Preventative Services Task Force stopped recommending routine PSA screening among asymptomatic patients to detect prostate cancer in 2012. They based their decision on a careful review of medical research, noting the screening didn't have much of an effect on mortality but could cause stress or invasive follow-up tests. Urologists generally opposed the decision. The USPSTF has since walked it back to, "Talk about the risks and benefits." I've looked at survey results for my state, and the numbers indicate a good proportion of men are told the benefits of a PSA but not any risks Patients are even less reasonable. If you tell somebody they have a tumor, they will now have a constant stress. If you say "cancer," they'll likely undergo expensive and potentially harmful treatment, even if "watch and wait" was a totally valid choice (e.g., slow-developing prostate cancer for very old men). Remember how Angelina Jolie had a double mastectomy after being told she had a good chance of developing breast cancer? That behavior will lead to a lot of unnecessary pain, debt, and lower-quality lives if it became normal. It'd be hard if not impossible to ask doctors don't share knowledge about a tumor with patients. But in some cases we intentionally ask them not to go looking for tumors because the expected value of a positive result is a negative impact.
  - nradov 5y ago
    Everyone gets cancer eventually, it's inevitable if you live long enough. There's no point in knowing that a small, slow growing tumor will kill you in 10 years if a heart attack is going to kill you in 5 years anyway. Knowing about the tumor just creates more psychological stress and potentially extra unnecessary medical treatments for no benefit.

pharmakom 5y ago

I don’t see radiology work decreasing either. Instead, I think it will serve more people but at lower cost per person. No one will skip medical services if they can afford them, but currently prices are high. Imagine a future where a radiologist serves 10x customers as before by leveraging smart technologies, for similar overall compensation.

carbocation 5y ago

I agree with you and with the sibling comment!

cloverich 5y ago

It is a good point but I think one part is a bit backward in an ironic way. One reason AI hasn't replaced radiologists is because radiologists are typically very good physicians, and specifically do not look at images in isolation but review the record, talk with the physicians, sometimes the patients too, etc. So its actually backwards (in some cases) -- AI struggles because it looks at the image in isolation, while the radiologist is looking at the patient more holistically.

koheripbal 5y ago
Radiologists don't talk to patients (usually), so there's no reason why AI cannot be given all the same patient data. ...although reading doctors' notes is probably a whole 'nother AI program.
- ska 5y ago
  > .although reading doctors' notes is probably a whole 'nother AI program. yep. One that also has a long history, and a lot of current players - an nobody has really good traction there either.

dx034 5y ago

But aren't we already over-diagnosing some cancers? Spotting more tiny tumors in unrelated images might do more harm (through procedures/treatment) than ignoring them. I'm not sure if we're really better off detecting every anomaly in someone's body.

vecter 5y ago
Why would we not want to know about a tumor in the body? I assume competent doctors will assess the risk of such a thing, but knowing about it is better than not.
- vharuck 5y ago
  Doctors will optimize for patient outcomes, usually by doing all they can. Sometimes, this doesn't scale well. For example, the US Preventative Services Task Force stopped recommending routine PSA screening among asymptomatic patients to detect prostate cancer in 2012. They based their decision on a careful review of medical research, noting the screening didn't have much of an effect on mortality but could cause stress or invasive follow-up tests. Urologists generally opposed the decision. The USPSTF has since walked it back to, "Talk about the risks and benefits." I've looked at survey results for my state, and the numbers indicate a good proportion of men are told the benefits of a PSA but not any risks Patients are even less reasonable. If you tell somebody they have a tumor, they will now have a constant stress. If you say "cancer," they'll likely undergo expensive and potentially harmful treatment, even if "watch and wait" was a totally valid choice (e.g., slow-developing prostate cancer for very old men). Remember how Angelina Jolie had a double mastectomy after being told she had a good chance of developing breast cancer? That behavior will lead to a lot of unnecessary pain, debt, and lower-quality lives if it became normal. It'd be hard if not impossible to ask doctors don't share knowledge about a tumor with patients. But in some cases we intentionally ask them not to go looking for tumors because the expected value of a positive result is a negative impact.
- nradov 5y ago
  Everyone gets cancer eventually, it's inevitable if you live long enough. There's no point in knowing that a small, slow growing tumor will kill you in 10 years if a heart attack is going to kill you in 5 years anyway. Knowing about the tumor just creates more psychological stress and potentially extra unnecessary medical treatments for no benefit.