AUTOMATED MEDICAL DIAGNOSIS

[Anonymous No.11128667]

My idea:

tl;dr: see pic related(1000 hours of MS paint)

>automate data gathering about someones health, so its cheap and fast enough for most people to do all of it once a month, or most of it, bloodwork, scans, diagnosed diseases, current treatments, etc in specialized clinics built for this purpose of data gathering.
This is the crucial step, if this is done, the rest will come more or less naturally. For this to be done, serious improvements in exam/tests automation are required, the less human work, the better.

>everyone will once a month grow their personal dataset with their health data
>those will feed into a huge central one
>this central database will be free and shared among researchers, with anonymized data for privacy concerns
>then these two processes will work in a cycle:
1) The central dataset will be used by researchers using big data technology to generate PROPER science and understanding of the human body and diseases, with data-based models that actually predict shit. Current medicine models are pathetic, vague and hand wavy, because of lack of data to be analyzed

2) The generated knowledge can be applied into the individual datasets, it would be possible to create end-user software that detects diseases and predicts future ones, then present treatments options or lifestyle choices in a user-friendly way, an obvious example: "your chloresterol is a little high, that can turn into coronary heart disease in X years, please consider Y lifestyle changes and/or Z medication treatment"

Then doctors will be a lot less necessary, mostly for niche cases or giving the greenlight on the recommended actions by the software, until full blown software autonomy is viable.

[Anonymous No.11128719]

>>11128667
Fixed with GIMP™ (boy was that a pain in the ass)

[Anonymous No.11128722]

>>11128667
You think you're the first one that thinks this shit, kiddo?

>automate data gathering about someones health, so its cheap and fast enough for most people to do all of it once a month
>technobabble technobabble
Automate how? And you think people are going to come to your shit clinic in the first place? Pregnant bitches don't even bother going to their fucking trimonthly checkups and you think you will gather enough information at your brainlet clinic?

What diseases are you going to screen? Who's gonna pay for all the shit?

Fuck off, idiot. I wonder why they let conceited dumbasses like you into universities, they should make it mandatory the idiots in the stem field go flip burgers and try to survive on their own for two years before letting them in so you have an useful skill after you go broke with your fantastical technomagical shit. Grow the fuck up.

[Anonymous No.11128729]

>>11128667
i like the general idea, but what goes into all these black boxes? that is the hard part. cadx is a thing for now but i expect diagnoses will be automated sooner or later.

[Anonymous No.11128748]

>>11128722
Well your 'tude isn't helping humanity, anon

[Anonymous No.11128778]

>>11128719
Rofl thanks for the laugh.
Yeah, solving the privacy issue probably wouldn't be easy, but thats just a side issue, the main idea still works regardless of the ethics involved.

>>11128722
>Automate how?
Automation technology is constantly improving, so I'm assuming that Some Day™ this cheap data gathering service will be a reality, then this process could be implemented.

>And you think people are going to come to your shit clinic in the first place?
Effortless general diagnostics and effective personalized treatment plans would be of the interest of anybody. Way easier than going to a bunch of specialist appointments and describing your symptoms subjectively to a dumb doctor that is barely listening.

>>11128729
>but what goes into all these black boxes? that is the hard part
Not really, most of it could be done with current technology to some degree of effectiveness IF we could gather the data.
>huge database containing all individual datasets
ETL/data warehousing processes that is already studied to death.

>process of gather knowledge from the data
State of the art pattern recognition technology, with this amount of standardized and complete data, even current tech would be pretty effective.

>software that present things to the user
Use the same tech above to compare the specific user data to the general knowledge base, then generate diagnostics/predictions/treatments and present to the user in a nice interface.

[Anonymous No.11128782]

>>11128667
Dumb. This is statistical modeling. You're not even close to what modern machine learning trained model are doing.

[Anonymous No.11128798]

>>11128782
>You're not even close to what modern machine learning trained model are doing.
Can you elaborate? Its just a general idea of "more/better data -> better models -> viabilize automated diagnostics", not about a specific training technology.

[Anonymous No.11128820]

Problem is that people don't understand their own symptoms, their significance, causality and oftentimes even miss obvious ones that an experienced clinician might catch, and that goes especially for the above average IQ hypochondriacs that do research before their appointments.

[Anonymous No.11128848]

>>11128820
Yeah, the theoretical best way to address this would be some method to quantify symptoms and store this data in the individual dataset, without subjective user input.
But of course this is way ahead than current technology, you can't really wear a smartwatch that measures how much pain you are feeling in X region of your body, or something like that.

[Anonymous No.11129652]

>>11128848
They are called symptoms because you can't fucking measure them, you brainlet autist.

Holy fuck. I thought you faggots bragged about being smart. How can you not understand the basic notion that shit people feel can't be measured.

[Anonymous No.11129718]

>Researchers gather knowledge from data using data science

Ok, you don't "gather knowledge", you create knowledge by inferring relationships - which is what data science does. You don't collect data using "data science" either, data science is how you analyse it.

>Yeah, the theoretical best way to address this would be some method to quantify symptoms and store this data in the individual dataset, without subjective user input.

But symptoms are subjective? They're subjective to the patient and they're subjective at the level of the clinician. Of course you can quantify some things (e.g. "Dark rash, spots) and collect measurements (blood pressure) etc, but as this guy put it:

>...people don't understand their own symptoms, their significance, causality ...

[Anonymous No.11130737]

>>11129652
>>11129718
>hurr symptons can't be measured because its subjective
No shit sherlock, thats what I just said. But my point is that in theory, the best way to work with this is scraping the clinical model where you ask the patient for their subjective symptons and develop new models based on objective data.
But thats the ideal case, not the viable one, currently we don't have the technology to replace a question "where does it hurt?" with a device that collects this data that give better answers than the patient answer to that question.

[Anonymous No.11130930]

>>11128667
MD here. Your first requirement would not be doable simply because current imagery and bloodwork are already too time consuming and expensive, (especially MRI) while "fast" nuclear imagery (scan/radio) has obvious health risks.
Doctors are hounded to do less tests on patients to maximize economic and time efficiency, extending these to healthy individuals simply isn't materially possible.

Mass data gathering is much more realistically done with recording of datas like steps, HR, tension, spO2, sleep, body temperature, etc...
Most things a smartwatch can record and which would help tremendously for screening, monitoring and predicting evolutions.

Also the vast majority of modern pathologies are chronic and don't require advanced tests either nor invasive treatments. In fact the current problem is in the majority of the case the patient who fails to observe his treatment and/or change his lifestyle.

The doctor seems to be the white whale of many people when it comes to automation (I would guess based on grudges or bad personal experience) and while I can see the profession being augmented in the next decades (especially since I'm a dev myself) there are too many chokepoints (being social, technological or economical) currently to be fully automated before decades if not more as all of these things require a technological proficiency which is simply absent among old people (again, making the majority of patients).

Finally doctors are the last link in the chain of healthcare, as I said earlier most cases today are chronic pathologies taking their root in the lifestyle.
If you want to really undercut doctors that's where the market opportunity is: personal automated health assistants, coaching and educating people to optimize their health.
You could imagine a Siri like assistant guiding purchases at the grocery store, cooking, exercising, etc... Moving the cognitive load from planning/choosing/recording to the AI so they don't need doctors at all.

[Anonymous No.11131543]

>>11128778
You're naive and a bit dumb. Like the other poster said, people can't even be assed to go to their pregnancy or diabetes check-ups every few months. You think they are going to voluntarily go get poked by needles and receive high doses of radiation so a machine can print out a paper that says they're fine?

This is only looks good for hypochondriacs, and to be honest you kinda sound like one.

[Anonymous No.11131602]

*Ahem*, threadly reminder
https://www.youtube.com/watch?v=J8nk7GrB-zs
14:10

[Anonymous No.11131640]

>>11131602
Irrelevant.