AI and medicine- the way ahead is here-

lousubcap

An insightful (to me) article about rapid advances in medicine and the use of AI.
https://www.theverge.com/c/22927811/medical-algorithm-explainer-sepsis-risk-watch  
Edit:  The intro to the above article-
""Artificial intelligence algorithms are everywhere in healthcare. They sort through patients' data to predict who will develop medical conditions like heart disease or diabetes, they help doctors figure out which people in an emergency room are the sickest, and they screen medical images to find evidence of diseases. But even as AI algorithms become more important to medicine, they're often invisible to people receiving care.""

nolaegghead

There's a fine line between risk assessment, diagnostic procedures and insurance reimbursement.   For profit hospitals, which is largely the US model, will never be as good as they could be.  The future of medicine is partly in prognostication if it's to be efficient.  Right now, the bulk of bad-outcome treatment is after symptoms rather than before.  We can avoid spending huge amounts of money and the poor outcomes with preventative care, which we sorely lack.

I'm just talking out my ass as an observer here, not vested in the system other than being just another paying patient.

Where medicine is too invasive for some people who perceive it "infringing on their freedum", we need to let them opt out.  There's a Samuel L Jackson quote for that, I believe.

JohnInCarolina

I’ve seen these systems in action and when used properly (to make recommendations to human experts) they are vastly superior to expert decision making alone.  The data sets are just so large and they’re able to make connections that are there but counter-intuitive to typical human understanding.

To be clear I’m just focused on the machine learning side of things.  Actual AI (ML plus machine-based decision making) I’m far more skeptical of.

nolaegghead

Machine learning is only as good as the quality and volume of material.  I seem to remember there are many obstacles to getting enough quality data to look at, in part because medical data is confidential by default.

JohnInCarolina

nolaegghead said:

Machine learning is only as good as the quality and volume of material.  I seem to remember there are many obstacles to getting enough quality data to look at, in part because medical data is confidential by default.

There has been an explosion of labeled data over the past decade.  That is a big part of what is driving the ML revolution.  Efforts to continue to label data and make it findable, accessible, interoperable and reusable are only continuing to increase.

nolaegghead

Yeah, it does take some work to label, add metadata and otherwise put the data into a digestible format.  I presume that is one of the largest expenses and where "big data" comes into play.

fishlessman

i dont see how this is going to work out til the doctors are all on the same system which looks to be along way off. absolutely non of my doctors have my full history over even just the last year, they ask me questions and i pull out test results out of the folder i carry with me to each visit. they take copies from it.  beth isreal faxes and scans, dana farber enters the data but has everyone locked out, pawtucket medical sets me up with doctors that moved on 2 years prior, the pharmacy fills up prescriptions ive never been given, and blue cross is just a nightmare to work with. last time getting bloodwork at quest, me and the woman taking blood were both reading the screen and guessing what tests i was going to get, they dont even use the same test codes. maybe things will speed up, i got an appointment for sleep apnea early dec, got the test done late january, i have a consultation next wednesday. thats 4 months, maybe they are sleeping over there

if the system doesnt work, then this AI thing is going to be dumber than a box of rocks

nolaegghead

fishlessman said:

i dont see how this is going to work out til the doctors are all on the same system which looks to be along way off. absolutely non of my doctors have my full history over even just the last year, they ask me questions and i pull out test results out of the folder i carry with me to each visit. they take copies from it.  beth isreal faxes and scans, dana farber enters the data but has everyone locked out, pawtucket medical sets me up with doctors that moved on 2 years prior, the pharmacy fills up prescriptions ive never been given, and blue cross is just a nightmare to work with. last time getting bloodwork at quest, me and the woman taking blood were both reading the screen and guessing what tests i was going to get, they dont even use the same test codes. maybe things will speed up, i got an appointment for sleep apnea early dec, got the test done late january, i have a consultation next wednesday. thats 4 months, maybe they are sleeping over there

if the system doesnt work, then this AI thing is going to be dumber than a box of rocks

I got a sleep study done in early Dec and just got my CPAP last week.   And I was complaining.

The diagnosis side of medicine using AI/machine learning doesn't really have anything to do with the dysfunctional medical infrastructure.  Other than it's less efficient not having a central medical records database.

JohnInCarolina

nolaegghead said:

fishlessman said:

i dont see how this is going to work out til the doctors are all on the same system which looks to be along way off. absolutely non of my doctors have my full history over even just the last year, they ask me questions and i pull out test results out of the folder i carry with me to each visit. they take copies from it.  beth isreal faxes and scans, dana farber enters the data but has everyone locked out, pawtucket medical sets me up with doctors that moved on 2 years prior, the pharmacy fills up prescriptions ive never been given, and blue cross is just a nightmare to work with. last time getting bloodwork at quest, me and the woman taking blood were both reading the screen and guessing what tests i was going to get, they dont even use the same test codes. maybe things will speed up, i got an appointment for sleep apnea early dec, got the test done late january, i have a consultation next wednesday. thats 4 months, maybe they are sleeping over there

if the system doesnt work, then this AI thing is going to be dumber than a box of rocks

I got a sleep study done in early Dec and just got my CPAP last week.   And I was complaining.

The diagnosis side of medicine using AI/machine learning doesn't really have anything to do with the dysfunctional medical infrastructure.  Other than it's less efficient not having a central medical records database.

This is exactly right.  The ML algorithm doesn't necessarily need to be fed all of your past data.  In principle one could be generated that only needs your demographics (age, weight, etc.), any medical test results, and a list of any symptoms.  

nolaegghead

JohnInCarolina said:

nolaegghead said:

fishlessman said:

i dont see how this is going to work out til the doctors are all on the same system which looks to be along way off. absolutely non of my doctors have my full history over even just the last year, they ask me questions and i pull out test results out of the folder i carry with me to each visit. they take copies from it.  beth isreal faxes and scans, dana farber enters the data but has everyone locked out, pawtucket medical sets me up with doctors that moved on 2 years prior, the pharmacy fills up prescriptions ive never been given, and blue cross is just a nightmare to work with. last time getting bloodwork at quest, me and the woman taking blood were both reading the screen and guessing what tests i was going to get, they dont even use the same test codes. maybe things will speed up, i got an appointment for sleep apnea early dec, got the test done late january, i have a consultation next wednesday. thats 4 months, maybe they are sleeping over there

if the system doesnt work, then this AI thing is going to be dumber than a box of rocks

I got a sleep study done in early Dec and just got my CPAP last week.   And I was complaining.

The diagnosis side of medicine using AI/machine learning doesn't really have anything to do with the dysfunctional medical infrastructure.  Other than it's less efficient not having a central medical records database.

This is exactly right.  The ML algorithm doesn't necessarily need to be fed all of your past data.  In principle one could be generated that only needs your demographics (age, weight, etc.), any medical test results, and a list of any symptoms.  

Yeah, too much data is almost a bad thing in some cases, not just in trying to be unbiased, but people will mine big data and write idiotic articles from faulty conclusions like "bingo contributes to higher incidence of lung cancer".  I mean, duh.

SamIAm2

Then there is the thing a GP's patient will say just as they are leaving the examination room, turning back toward the doctor to speak that one last thing revealing more information than the patient may realize and is the final bit needed to make a proper diagnosis.  

JohnInCarolina

SamIAm2 said:

Then there is the thing a GP's patient will say just as they are leaving the examination room, turning back toward the doctor to speak that one last thing revealing more information than the patient may realize and is the final bit needed to make a proper diagnosis.  

That can certainly happen too, but that would occur without the ML algorithm.  As I wrote up above, when these things work best they make suggestions - for a diagnosis or an additional test to run - that the GP wouldn’t have thought of on their own.  In the end they’ll end up saving a lot of lives, just watch.