Medicine's Machine Learning Problem • Hacker News Board

Re: Medicine's Machine Learning Problem

#101

Post by Gatsky » Sat, Jan 09, 2021, 2:01 AM UTC

tpfour wrote: post #100

gdebel wrote: post #14

I'm an eye surgeon and self-taught machine learning practitioner, I started to learn Python in 2016 when the deep learning hype was at his highest. After 3 years of research, playing with datasets, extracting and cleaning data from EMR and from different machines, I not sure that the biggest problem with the so-called "AI" is the inequalities that it can induce ; it is rather, is it useful at all ? This is a little b…

Tangential: I've always had an interest for medicine. A decade or so into software dev and I'm not convinced I want to stay. I've heard a lot of doctors say they wouldn't do it all over again. I'm wondering if med school makes any sense. Maybe my idea of medicine is too influenced by Hollywood, but it does look very interesting. Would you study medicine again given the chance to start over?

This depends greatly on where you will be studying, where you intend to practice, and your matrix of personal and financial dependencies.

Working conditions in medicine are universally poor, frankly. The flexibility during education and training is close to zero. You probably will need to move somewhere else as part of training or work. This can uproot your life and family depending on obligations. Some fields have very prolonged training periods and are very competitive (eg surgical subspecialties). There is a lot of bullshit in medicine, due to heirarchies in large institutions, rigid and nonsensical regulations, issues with billing etc... you have to have the personality to stomach this. Conversely, there is also the fact that when you are in the room with the patient and the door is closed, it is just you trying to help another person, despite everything. This is sacred, and can be intensely rewarding even in dire circumstances, an exercise in mutual gratitude you just won't find elsewhere.

The other thing I would say is that some fields in medicine are probably not something you should spend your whole life doing, at least not full time. It is just too emotionally exhausting, no matter what you do to deal with it. Not saying it is impossible to do it for 40 years, but it is a risk, you definitely sustain some kind of damage. I suspect this is responsible for 2/3rds of the seemingly outrageous negligence/misconduct cases you hear about.

Re: Medicine's Machine Learning Problem

#102

Post by dnautics » Sat, Jan 09, 2021, 7:18 AM UTC

JimTheMan wrote: post #85

dnautics wrote: post #21

Earlier quoted context omitted.

How would you know without the data? There are plenty of medical conditions with wildly divergent rates and pathophysiologies based on human genetics.

I don't think the burden is on us to prove a negative. It'd be way better to think of it in terms of genetic markers instead of races. Race in general practice is a social construct based on colour, appearance and culture. It's a level of abstraction away from the actual genetic data that we don't need with our level of technology. We could dig right into the genes and throw away our outdated notions. That's where th…

"Race in general practice is a social construct based on colour, appearance and culture"

That is a very dangerous lie. If you are a doctor I pray for your patients that you are not taking a colorblind approach because their lives could be at stake. It just so happens that the way people look correlates with the diseases they get and sometimes don't get. Sometimes you can't sequence the person's genome because it's too time consuming. Sometimes we don't even fucking know which genes are responsible.

Your example of huntington's is quite frankly the outlier, and a facile example (you know it is because that's what they teach in 8th grade biology class). And even that's a shitty example because the severity only roughly correlates with number of tandem repeats. Any given person can tolerate a load of misfolded huntingtin, and their tolerance is probably governed by a raft of other conditions, that are familial and do track with race (venezuelans are for some reason more sensitive to huntingtin repeats). Other conditions, too, transthyretin amyloidosis is more common among japanese and finns and really rare among africans (except for one variant that causes congestive heart failure; but in finns and japanese it presents as liver failure), etc etc etc.

Re: Medicine's Machine Learning Problem

#103

Post by singularity2001 » Sat, Jan 09, 2021, 8:52 AM UTC

gdebel wrote: post #14

I'm an eye surgeon and self-taught machine learning practitioner, I started to learn Python in 2016 when the deep learning hype was at his highest. After 3 years of research, playing with datasets, extracting and cleaning data from EMR and from different machines, I not sure that the biggest problem with the so-called "AI" is the inequalities that it can induce ; it is rather, is it useful at all ? This is a little b…

   AI is not fundamentally different from a simple linear regression

it is fundamentally different because it is complex nonlinear regression (unless you are using a one layer linear network which no one does)

Re: Medicine's Machine Learning Problem

#104

Post by JimTheMan » Sat, Jan 09, 2021, 10:43 AM UTC

dnautics wrote: post #102

JimTheMan wrote: post #85

Earlier quoted context omitted.

I don't think the burden is on us to prove a negative. It'd be way better to think of it in terms of genetic markers instead of races. Race in general practice is a social construct based on colour, appearance and culture. It's a level of abstraction away from the actual genetic data that we don't need with our level of technology. We could dig right into the genes and throw away our outdated notions. That's where th…

"Race in general practice is a social construct based on colour, appearance and culture" That is a very dangerous lie. If you are a doctor I pray for your patients that you are not taking a colorblind approach because their lives could be at stake. It just so happens that the way people look correlates with the diseases they get and sometimes don't get. Sometimes you can't sequence the person's genome because it's to…

I agree with you, we shouldn't ignore the very real strong correlation for use in medicine right now and in the future.

I reckon I've thrown you with the general practice comment, which I intended to mean "everyday use" and not as a "general practitioner of medicine".

Clearly this is something you feel very strongly about but I think you'd get further with less belligerence. I feel like I've unknowingly wandered into the office of reviewer #2 and he's flown off the handle at something I am unwilling to match the level of aggression on.

Re: Medicine's Machine Learning Problem

#105

Post by dnautics » Sat, Jan 09, 2021, 5:31 PM UTC

JimTheMan wrote: post #104

dnautics wrote: post #102

Earlier quoted context omitted.

"Race in general practice is a social construct based on colour, appearance and culture" That is a very dangerous lie. If you are a doctor I pray for your patients that you are not taking a colorblind approach because their lives could be at stake. It just so happens that the way people look correlates with the diseases they get and sometimes don't get. Sometimes you can't sequence the person's genome because it's to…

I agree with you, we shouldn't ignore the very real strong correlation for use in medicine right now and in the future. I reckon I've thrown you with the general practice comment, which I intended to mean "everyday use" and not as a "general practitioner of medicine". Clearly this is something you feel very strongly about but I think you'd get further with less belligerence. I feel like I've unknowingly wandered into…

The aggression is not to convince you. It's to help point out how dangerous your naive reading is to other people. Don't take it personally.

Re: Medicine's Machine Learning Problem

#106

Post by JimTheMan » Sat, Jan 09, 2021, 10:10 PM UTC

dnautics wrote: post #105

JimTheMan wrote: post #104

Earlier quoted context omitted.

I agree with you, we shouldn't ignore the very real strong correlation for use in medicine right now and in the future. I reckon I've thrown you with the general practice comment, which I intended to mean "everyday use" and not as a "general practitioner of medicine". Clearly this is something you feel very strongly about but I think you'd get further with less belligerence. I feel like I've unknowingly wandered into…

The aggression is not to convince you. It's to help point out how dangerous your naive reading is to other people. Don't take it personally.

Good luck navigating the world as an asshole. Let me know if it pays off for you.

I've taken it very personally. I suspect others in your life do too.

Re: Medicine's Machine Learning Problem

#107

Post by tpfour » Sun, Jan 10, 2021, 1:11 AM UTC

Gatsky wrote: post #101

tpfour wrote: post #100

Earlier quoted context omitted.

Tangential: I've always had an interest for medicine. A decade or so into software dev and I'm not convinced I want to stay. I've heard a lot of doctors say they wouldn't do it all over again. I'm wondering if med school makes any sense. Maybe my idea of medicine is too influenced by Hollywood, but it does look very interesting. Would you study medicine again given the chance to start over?

This depends greatly on where you will be studying, where you intend to practice, and your matrix of personal and financial dependencies. Working conditions in medicine are universally poor, frankly. The flexibility during education and training is close to zero. You probably will need to move somewhere else as part of training or work. This can uproot your life and family depending on obligations. Some fields have v…

Thanks for the reply. A part of me believes it would be worth going through med school just to get to "save" one life. I just don't get that gratitude feeling as a developer. I mean sure, you can work on software that could be used by millions of people, software _does_ change and affect the world, but it does not really compare to medicine. A git commit can be immensely important but there's something about medicine that makes it very appealing. On the other hand, computers do as they're told and rarely complain, stink, whine ;)

I am strongly attracted but I feel maybe it's mostly illusory.

Re: Medicine's Machine Learning Problem

#108

Post by n3ur0n » Sun, Jan 10, 2021, 4:41 PM UTC

BlueTemplar wrote: post #93

n3ur0n wrote: post #52

Earlier quoted context omitted.

I do respect your experience and take on the matter, however, let's replace this statement: "I'm an eye surgeon and self-taught machine learning practitioner, I started to learn Python in 2016 when the deep learning hype was at his highest." with: I'm a [machine learning researcher] and self-taught [ophthalmologist], I started to learn [ophthalmology] in 2016 when the [clinical medicine] hype was at his highest. In t…

I have barely any biology/medecine nor Machine Learning knowledge (though some physics, maths, programming), yet I might have to do an internship in the field of ML applied to leukocyte classification, where would you recommend to start ?

Depends on the scope of the project. Would the goal be to come up with a better algorithm for cell classification based on histological images? Or to apply an existing algorithm to a new dataset?

The former would be quite difficult without much background in ML/Computer Vision (you would have to spend some time self-teaching basics of ML/Deep Learning and the pre-reqs for those — Basic Linear Algebra and Probability).

The latter is doable. I would recommend a very hands on approach. Pick some computer vision object classification tutorials and code them up (using a high level library). Make a mind map of the concepts and look them up as and when you’re unclear about a concept. Then move on to replicating some well cited, peer reviewed papers. Often papers will have their code on GitHub. Try and relocate their results on their dataset. After this you would have the basic working knowledge to modify the algorithm slightly for your specific use case.

Re: Medicine's Machine Learning Problem

#109

Post by dnautics » Mon, Jan 11, 2021, 5:14 AM UTC

JimTheMan wrote: post #106

dnautics wrote: post #105

Earlier quoted context omitted.

The aggression is not to convince you. It's to help point out how dangerous your naive reading is to other people. Don't take it personally.

Good luck navigating the world as an asshole. Let me know if it pays off for you. I've taken it very personally. I suspect others in your life do too.

I'm doing pretty good actually. Hope that ruins your day.

Re: Medicine's Machine Learning Problem

#110

Post by BlueTemplar » Mon, Jan 11, 2021, 9:46 AM UTC

n3ur0n wrote: post #108

BlueTemplar wrote: post #93

Earlier quoted context omitted.

I have barely any biology/medecine nor Machine Learning knowledge (though some physics, maths, programming), yet I might have to do an internship in the field of ML applied to leukocyte classification, where would you recommend to start ?

Depends on the scope of the project. Would the goal be to come up with a better algorithm for cell classification based on histological images? Or to apply an existing algorithm to a new dataset? The former would be quite difficult without much background in ML/Computer Vision (you would have to spend some time self-teaching basics of ML/Deep Learning and the pre-reqs for those — Basic Linear Algebra and Probability)…

The data in the database comes from a bidimensional matrix (LMNE) where leucocytes are classified on resistivity on one axis and light absorption (?) on the other. (I wonder how they managed the separation by absorption... indirectly via centrifugation ?) So I guess not really histological ?

Looks like it's a new model, I have no idea if they already have any ML models yet. There's also some database work.

I'm finishing a Masters degree in Computational Physics, so Linear Algebra and Probability shouldn't be an issue. (We also have an Image Processing and Analysis course.) I guess that's why they contacted us despite the fact that we don't have any ML training ?

Yeah, this is basically what I thought to do, but thank you for your advice !