TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Deep learning on electronic medical records is doomed to fail

227 pointsby brileeabout 3 years ago

44 comments

ZeroCool2uabout 3 years ago
Having worked with data from EMR systems and having worked at a large EMR software development shop myself, and now using deep learning at work quite a bit for the past few years, I&#x27;m inclined to agree.<p>This title is somewhat click bait though, because the fault is really with EMR systems and (esp) the American Healthcare system, not deep learning.<p>The entire system is designed around billing and decisions are made my hospital and insurance executives that are generally not technical. There is no incentive to clean up the system or work on a well structured open protocol for interop the same way there is in say banking. Plus, the author gives some good examples like pulse ox%, doctors and nurses are not at all concerned with or trained to record data in a way that makes sense to use programmatically. They&#x27;re typically thinking only as if they&#x27;re recording it for another human to read.<p>Deep learning could probably be quite useful in the medical field, but we won&#x27;t know until someone comes along and disrupts the system top to bottom similar to how Tesla has done with not only manufacturing, but the sales process and shirking the dealership model. This would probably look something like Forward[1], but with a crazy amount of funding, so that insurance companies and billing codes could be ignored entirely.<p>[1]: <a href="https:&#x2F;&#x2F;goforward.com&#x2F;p&#x2F;home" rel="nofollow">https:&#x2F;&#x2F;goforward.com&#x2F;p&#x2F;home</a>
评论 #30766879 未加载
评论 #30769599 未加载
评论 #30767186 未加载
评论 #30769825 未加载
评论 #30768975 未加载
评论 #30767282 未加载
评论 #30767339 未加载
评论 #30767266 未加载
评论 #30767207 未加载
评论 #30775445 未加载
评论 #30769762 未加载
AutumnCurtainabout 3 years ago
The first point he raises is the most critical by far. The silverbacks of the industry deliberately stymie efforts for true interoperability because it goes directly against their primary goal, which is forcing everyone into their platform. Epic in particular has zero intention of allowing anyone else to take their market share by enabling easy sharing of data across platforms. It&#x27;s far better for them from a business perspective to make interfacing so unreasonably difficult that you are forced to implement their full suite of applications, at which point they hold your organization&#x27;s data hostage to induce other orgs to do the same. The larger their ecosystem grows, the less they need to worry about interoperability - improving patients&#x27; outcomes is not even an afterthought. Their vision of population health reporting is one in which every major healthcare org has been trapped inside their walled garden.
评论 #30766817 未加载
评论 #30766810 未加载
评论 #30768305 未加载
评论 #30768025 未加载
boasabout 3 years ago
&gt; Life would be simpler if only these hospitals could set aside their arrogance and just go with the recommended workflow!<p>This would be like asking programmers to standardize on the recommended programming language.<p>we would love to just use the recommended workflow, if it worked for our hospital. There are differences in the patients, doctors, local regulations, existing systems, etc between hospitals.<p>Patients: Top cancer hospital does a lot of clinical trials, so some of the forms require you to fill out clinical trial information for every patient. In a maternity ward, it would not be appropriate to ask about clinical trials for every patient.<p>Doctors: Hospitals are staffed differently. If the hospital has residents, some of the work can be delegated to residents. If not, someone else has to do it. The workflow needs to account for who is actually available to do the work.<p>Local regulations: Medicine is highly regulated, and each state and hospital has its own rules.<p>Existing systems: Hospital computer systems have been around for decades, and usually it&#x27;s not possible to migrate everything to a new system, so the new system needs to integrate with the old systems that couldn&#x27;t be upgraded.
评论 #30766886 未加载
idohabout 3 years ago
I happen to know a lot of doctors, including, as an example, an OBGYN. As it was explained to me, for vaginal births:<p>- at some point someone, without evidence, speculated that cervix dilation should proceed along some curve<p>- cervix dilation is actually measured by hand - literally inserting fingers and having the doctor practice &quot;so many fingers = so many centimeters&quot;. There&#x27;s plastic sheets with holes in them so they can practice measuring the size of holes with fingers.<p>- the OBGYN knows that the cervix dilation curve should look like, and kinda sorta maps their hand readings to what it should look like<p>- the OBGYN has a general sense as to how labor is going, and will game the cervix dilation stats to match their expectation, e.g. if labor is going well but the cervix hasn&#x27;t dilated then they&#x27;ll kinda sorta report progress anyway<p>Anyway, given the above it seems like the data around cervix dilation is suspect - the measurements are fitted to what the curve should look like, and then the data matches the curve, and that makes people more confident in the curve, and so on.<p>The point is, can you really apply ML to the EMR of cervix dilation? Does it make sense, could you really draw conclusions from this?
评论 #30783257 未加载
评论 #30771266 未加载
cptajabout 3 years ago
&gt;EMR software is widely hated by the nurses and doctors who have to use it. It’s slow, bloated, nonintuitive, requires workarounds, etc. etc. etc.. The root of this evil is that every hospital brings its own conceited and byzantine patchwork of procedures, checks, and rituals to the table.<p>You just described every admin software for every industry I&#x27;ve worked on.<p>The problem is individual orgs dictating software architecture. When each purchase is in the millions, you accommodate every whim no matter how absurd... and then you end up with these bloated, messy systems.<p>For software systems to REALLY, shockingly improve efficiency in an organization, all the processes in the org need to change to accommodate a new overarching system design. Tailoring software to mirror legacy processes defeats the purpose almost entirely.<p>I think there is a truly absurd competitive advantage in doing this right but you seldom see enough leverage to completely overhaul every department in order to implement software admin systems.
评论 #30771587 未加载
ialyosabout 3 years ago
The article is simply wrong.<p>I know this because I worked as an ML engineer at an extremely successful company that automated medical coding using deep learning.<p>The confusion stems from conflating a &quot;perfect solution&quot; with a &quot;human augmented&quot; one.<p>90% of coding cases are trivial, have low value and can be done by a model. 10% are really subtle and need human expertise.<p>That&#x27;s fine. You can make a billion dollar company on low hanging fruit. I think it&#x27;s best not to conflate the perfect solution with a very good solution.
评论 #30770315 未加载
评论 #30767610 未加载
评论 #30769680 未加载
fnbrabout 3 years ago
I agree with the conclusion. This is totally unsurprising to me as a ML engineer. If you put garbage data into the model, you get garbage predictions. That doesn’t strike me as particularly novel. The same is true for cooking, after all.<p>However- this has been truly shocking to all of the non-technical stakeholders I’ve worked with. They take the stance that any large amount of data can be used to do ML on, presumably because they don’t know too much about what doing ML is like.<p>So I’m convinced the author is right, and I’m also convinced that there will be many attempts to use ML on EMRs.
评论 #30766872 未加载
评论 #30766649 未加载
deltarholamdaabout 3 years ago
While a lot of the post is good info, there are some upsides, if we can get the EHR situation worked out.<p>Many years ago, prior to anything like ML, Canada figured out that cystic fibrosis patients whose weight is higher than 50th percentile, had significantly better lung function. Nobody really understood why, but the correlation was so strong (.85 or something like that) it could not be ignored. Treatment protocols for CF changed to encourage weight gain, and lifespan outcomes have steadily improved over the years.<p>What other oddball correlations are hiding in the depths of bloodwork, weight&#x2F;height, etc. for patients? We&#x27;ve teased out all the easy ones, the ones that are left are combinations nobody thought to even measure.<p>Regarding the EHR debacle, I&#x27;m optimistic that something could be worked out as a standard and implemented across the board. Expensive? Sure, but it&#x27;s an investment that pays off pretty quickly, I think.
评论 #30770297 未加载
paulsutterabout 3 years ago
Better title, &quot;ML on EMRs is very difficult&quot;. The article is much more reasonable than the clickbait title:<p>&quot;I don’t think deep learning models on EMRs are going to be useful any time soon .. Clinical expertise is absolutely necessary to ask the right questions, to set up the inputs to the model, and to sift through the findings. Following that, clinical research will be necessary to validate the discovery&quot;
wise0wlabout 3 years ago
I previously worked for a startup that did denormalization of healthcare data for the ostensible purpose of data &quot;freedom&quot; and interoperability, with a future focus on ML funsies. All the issues we had (besides ones we created) were around the healthcare providers fear of pissing off EPIC, Cerner, Meditech, Allscripts etc. They didn&#x27;t <i>like</i> their EMRs---in fact they often hated them. The fact is though that there really is no viable alternative, and the data is kept behind a gate and essentially &quot;owned&quot; by the EMR. FHIR was supposed to solve the interoperability bit, but all the EMRs would still own the data, and their lawyers aren&#x27;t keen to share.
wutbrodoabout 3 years ago
Leaving aside the well-known dumpster fire of the healthcare system&#x27;s operational incompetence, none of these are particularly novel, most of them are the author discovering basic best practices in statistics, and they don&#x27;t come close to implying &quot;DL for EMR is doomed&quot;.<p>A health outcome is correlated with age? Ya don&#x27;t say.... This is only a problem that literally every single economist and sociologist checks for _first_. Agents in causal graphs respond to their inputs? Shocker, that&#x27;s only.... The definition of an agent.<p>There are a ton of applications out there where a fairly naive model built by someone who took a Pytorch Coursera course can provide decent improvements. Healthcare is not one of them, and nobody serious ever thought it was. Bringing modern learning tools into healthcare is going to require a lot of smart people who know what they&#x27;re doing, working cautiously to introduce these improvements into an operational quagmire with high stakes.<p>But this article reads like: &quot;I tried making healthcare &#x27;smart&#x27; in a Jupyter notebook over a weekend and it didn&#x27;t work: the effort is doomed&quot;.
jesseryoungabout 3 years ago
I&#x27;ve worked in healthcare IT for my entire professional career - It&#x27;s A LOT more complicated than most people think. For the last 5 years I&#x27;ve focused on the data side of healthcare and I think that deep learning is 100% possible - it&#x27;s just not achievable by a single person and it&#x27;s likely VERY expensive. There are so many facets to healthcare data that&#x27;s it&#x27;s just impossible for a single individual to achieve something meaningful by themselves without the help of teams of doctors, data analysts, data engineers and data scientists. Just dealing with data quality issues (such as the ones called out in this essay) require a team of people to determine if metrics you are trying to measure are legit or not.<p>On billing: I&#x27;m convinced that the primary reason why healthcare (at least in the US) is so complex - is because of the dichotomy of saving people at all costs, while doing so fiscally responsibly. It is fairly common for large healthcare organizations to have ACTING doctors in their c-suite, who&#x27;s primary goal is not to make money - it&#x27;s to save lives. The people who care about saving money, reducing cost and increasing efficiency have no control over the organization. I&#x27;m not saying this is a bad thing, but IMO it&#x27;s the largest contributing factor as to why healthcare billing is so complex, and healthcare costs get as high as they do (at least in the US).
methehackabout 3 years ago
<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Carte_Vitale" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Carte_Vitale</a><p>France&#x27;s system is universal and private (private docs, insurance, and reference pricing -- a service has a single fee across coherently regulated payers). They have had a standardized medical record for 20+ years. One system and one medical record for everyone. WHO ranked #1 healthcare in the world (to US ~40th) at 1&#x2F;2 the cost per capita of US healthcare. This is catastrophic legislative failure for a problem largely solved by lots of other people around the world.<p>The legislative failure has created vast administrative overhead (10, or more, staff per doc at a hospital) and corrupt insurance companies. When an insurance company has to pay a claim, they call it a &quot;medical loss&quot; (they had the money and they lost it). They make their money on poor service and deceit -- charging wildly different prices for the same product where they can get away with it). In France, an insurance company, by law, has to pay a claim to a practice in a few days. Imagine the decreased capital needs for running a medical practice or a hospital.<p>The hospitals are not blameless in all this, but the heart of it is the payer system&#x2F;s.
评论 #30768282 未加载
评论 #30767822 未加载
评论 #30768807 未加载
评论 #30767409 未加载
评论 #30767527 未加载
评论 #30768168 未加载
评论 #30767771 未加载
评论 #30768818 未加载
评论 #30767898 未加载
staredabout 3 years ago
Most of the points are naive and confuse predictive power with interpretability (the latter is harder). For example:<p>&gt; Did you know that a blood oxygen saturation of 0% is highly correlated with healthy outcomes? No, I didn’t get the percentage backwards. A 0% reading is what you get when the nurse looks at you and decides you’re too obviously healthy to bother with putting the pulse oximeter on your finger. The empty field value gets saved as a 0, of course.<p>Well, statistical methods (no matter if classical, Bayesian, Deep Learning, or anything, as long as it goes beyond linear methods) will perfectly capture the special case of 0% and predict accordingly. These methods are free of our biases and will take consistent approaches (e.g. empty columns, columns that de facto mean something else, common typos, etc).<p>Sure, interpretability is problematic, and we often need to consider the knowledge of physicians and medical institutions rather than be based on raw data.
midjjiabout 3 years ago
Deep learning has great potential in medicine, in particular in radiology and tissue classification. Creating the datasets from scratch will take decades of careful deliberate highly costly effort however, and the current crap hospitals call records is utterly useless. It will truly have to be a bottom up approach, and in the process systematic studies to actually verify a lot of bullshit medical ideas will also have to be done. Basic questions like how many kinds of tissues are there have very dubious answers which are known to be coarse approximations, and some diseases are specifically deviation from the approximations. Its probably not quite as bad as linguistics where, but its really bad. Once datasets with millions of people followed and tested regularly throughout their lives for the specific purpose of generating the dataset, are available for training, it will be quite good. Shame we wont live to see it.
adultSwimabout 3 years ago
These systems are primarily designed to support clinical care. Research is an after thought. Health care systems will have to decide it&#x27;s a top level priority.<p>There has been modest progress through wider efforts:<p>- Standard vocabularies, eg LOINC codes for different kinds of lab tests<p>- Mappings between vocabularies, eg OMOP<p>- Semantically rich vocabularies, eg OBO&#x27;s OBI
suifbwishabout 3 years ago
Where deep learning will prevail in medical is in predictive medicine. When it finally becomes common to have personal genomic data available during checkups, the machine learning will be able to guess&#x2F;order tests for individuals based on their age, specific genes they are carrying, known life age that diseases onset ect, diseases for that area. It will also be able to look across the population and detect statistical patterns in geographic incidences of diseases with environmental causes which occur outside of the normal expected distribution or in hotspots.<p>The important thing to remember about AI is you need reliable data to train it as well as reliable data to test it. If you don’t the FDA will not allow you to employ it medically.
评论 #30770398 未加载
Dan_-about 3 years ago
Setting aside the sensationalist headline, the entire premise of the article is flawed. It&#x27;s a case of not even being wrong. Of course you&#x27;re going to get spurious results using poor data.<p>The author&#x27;s attempt at using structured EMR data is the root cause. We have found that structured data, which the author attempted to use, is at best 35% accurate. Sure it&#x27;s better than claims, but it does not reach the level of quality necessary to inform clinical decision-making. The reason for this is that almost everything clinically relevant is captured in freeform text fields--clinical notes. To build proper models from information in EMRs, you have to start with processing the narrative data, which is a hard problem.<p>Training models to interpret clinical notes requires clinical expertise. Clinicians record facts differently in different locations, and there are many different ways to say the same things, and sometimes they skip underlying facts because some other fact implies the rest. Different specialties record things differently too. You really cannot just throw some data into a notebook and hope it works. Even with clinician input, we still find that high quality results require ensemble models with multiple techniques; plain NLP doesn&#x27;t work either.<p>Take for example, non-alcoholic steatohepatitis (NASH), the leading cause of liver failure requiring liver transplant. NASH is a complication of non-alcoholic fatty liver disease, in which your liver has unusually large deposits of fat. NAFLD is not coded in structured data. To identify it from unstructured data, you have to extract concepts related to liver cancer, pre-diabetes, alcohol use, liver fibrosis, cirrhosis, jaundice, fatigue, and loss of appetite. To make a long story short, you cannot do these things using structured data or naive NLP approaches. F1 is zero.<p>So maybe his point, &quot;Data encodes clinical expertise&quot; is worthwhile, but the rest of the article...not so much.<p>Source: My company, Verantos <a href="https:&#x2F;&#x2F;verantos.com" rel="nofollow">https:&#x2F;&#x2F;verantos.com</a> , specializes in the generation of high-validity evidence from data we abstract from EHRs using machine techniques.
评论 #30767473 未加载
mercurywellsabout 3 years ago
Is part of the solution going to be having ML figure out what data might be missing to make a better conclusion, then explicitly asking the patient (or the person gathering data from the patient) for that missing data or a clarification?
axg11about 3 years ago
The author is spot on - the current approach of applying ML to flawed and inconsistent data is doomed. However, on the bright side, I think they also highlight a possible path for data scientists to bring value to healthcare. All of the examples of relationships and correlations were spurious, but if you keep digging you will eventually unearth interesting relationships. A minority of these relationships will be useful for improving hospital administration. Examining the relationships might not necessarily improve health outcomes but there is a real chance to make hospitals more efficient using data science.
ltbarcly3about 3 years ago
If there is no meaningful statistical information in medical records, then we should stop keeping them? By hypothesis a doctor who opens a medical record can&#x27;t gain any meaningful information from it for the same reasons listed in the article. I think this is sufficient to demonstrate the article is incorrect, there is significant information in medical records, and therefore it will be possible to train a model which can reproduce missing information to some degree.
评论 #30775144 未加载
RcouF1uZ4gsCabout 3 years ago
&gt; The answer turns out to be rather mundane. Pap smears are not recommended for women older than 65, and heart failure onset is typically around age 65. The pap smear just turns out to be a really good age bucketing signal to the model.<p>Pap smears are also a really good sex bucketing signal, and there are lots of diseases that are more prevalent in one sex, so I would expect Pap smears to be correlated and anti-correlated with a lot of other diseases as well.
jcimsabout 3 years ago
I&#x27;m running through a similar situation in risk management. There is so much domain and institutional knowledge that&#x27;s encoded in rules that its nearly impossible to reason about them in a generic way. Add in operational that is also rife with data quality and coverage issues and it becomes quite difficult.<p>I think we have a term for this in both areas: garbage in, garbage out.
aliu22about 3 years ago
&quot;Doomed to fail&quot; is too strong IMO.<p>All the problems the author brings up, while legitimate, are being worked on.<p>For example, on the interoperability front, TEFCA is making big strides on government-mandated nationwide interoperability: <a href="https:&#x2F;&#x2F;www.healthit.gov&#x2F;topic&#x2F;interoperability&#x2F;trusted-exchange-framework-and-common-agreement-tefca" rel="nofollow">https:&#x2F;&#x2F;www.healthit.gov&#x2F;topic&#x2F;interoperability&#x2F;trusted-exch...</a><p>&gt; In January 2022, ONC and the RCE announced the publication of the Trusted Exchange Framework and the Common Agreement (TEFCA). Entities will soon be able to apply and be designated as Qualified Health Information Networks.<p>Google also has made significant strides on deep learning on EHRs: <a href="https:&#x2F;&#x2F;ai.googleblog.com&#x2F;2018&#x2F;05&#x2F;deep-learning-for-electronic-health.html" rel="nofollow">https:&#x2F;&#x2F;ai.googleblog.com&#x2F;2018&#x2F;05&#x2F;deep-learning-for-electron...</a>
lumostabout 3 years ago
The challenge with ML and DL systems is that it&#x27;s difficult to know a-priori what will and won&#x27;t work. The math would indicate that there is nothing a suitable DL system cannot learn, however in practice certain neural architectures can only learn certain inputs. The cycle time to develop a new system is long, and data is unfortunately scarce. Developing a DL system to solve a problem involves guesswork as to the impact of innovating on any one of dozens of components.<p>Which is to say, it is and will be difficult to create a business based on applying a novel DL method to a particular problem space. We&#x27;re seeing a consistent trend that focusing on the other aspects of the problem such as data, tooling, or end to end services tends to be much more successful.
lepapillonabout 3 years ago
I don&#x27;t really want to comment on whether or not DL is doomed to fail on the EMR, but coming from an EMR background, I can say he lays out very accurate points. I particularly like how he explains #3 concisely, and it&#x27;s a point I use to criticize the private healthcare system.<p>The continual war between hospitals having to opportunistically charge for their services vs. the insurance industry having to take a default stance of deflection creates the massive, meaty layer of coding and billing waste. Thousands upon thousands of jobs exist just for this purpose, and I think any inefficiency in a single-payer system is more than offset by getting rid of that layer and everyone benefits.
elijabout 3 years ago
I agree with all points with respect to static EMRs including NLP efforts.<p>The cadence, uniformity and alignment of event stores to underlying pathways does pose an opportunity (but there&#x27;s so little research and I suspect Brian didn&#x27;t have access to this space). An EMR is a projection&#x2F;point in time snapsot of these events (basically API calls). There&#x27;s also inherent natural labels because inferences are evaluated against an actual pathway.<p>Specifically speaking about intra-episodic inferences -- longitudinal predictions (over several episodes -- or a patient life time) becomes wildly inaccurate. There&#x27;s a lot of research to demonstrate this no matter which models are used.
ilakshabout 3 years ago
We need either A) competent, highly technologically sophisticated government (which seems very far away obviously) or B) something really similar to take it&#x27;s place.<p>So much of what government does (aside from the bombings etc.) is really about providing and enforcing a framework for people to work together. And in this era that needs to be a high tech framework.<p>Actually, it needs to not only be very high tech, but also very cutting edge, decentralized, sufficiently holistic but also flexible enough to evolve.<p>Which is incredibly hard, and we probably will not get due to greed, stupidity and politics, and that may be the actual reason that human civilization is superceded by AI civilization.
dekhnabout 3 years ago
There is something I&#x27;ve been developing over the years which I call &quot;Konerding&#x27;s Empirical Observation #7&quot;: every attempt to improve health care with technology in the US will only every increase the cost while also decreasing the quality of service (on average).<p>This goes hand in hand with Konerding&#x27;s 3rd empirical observation, which is that there is a never ending supply of &quot;machine learning geniuses&quot; who are naive about how health care works in the US, and spend some 20 years learning how hard it is to actually do anything actionable with health care data.
jrapdx3about 3 years ago
I&#x27;m an American physician. I&#x27;ve practiced in a couple of specialties and know the insurance billing drill (or should I say game) pretty well.<p>Medical record systems, as other comments point out, have been constructed for the benefit of administrators, recording clinical data is mainly to support administrative needs.<p>In reality manifestations of a given illness vary <i>continuously</i> across a wide spectrum. It means patients with the same diagnosis have differing sets of symptoms and course of illness. As I like to say it, &quot;no two patients have exactly the same disease.&quot;<p>Official disease classification schemes embody the &quot;splitter&quot; model which attempts to fit continuous data into discrete categories. Of course sharp-edged distinctions suit purposes like billing and other management operations.<p>However diagnosis is often ambiguous, mixed or multiple, but no matter what categorical diagnoses must be assigned. Unsurprisingly doctors are biased to select the choices with the greatest reimbursement, and &quot;stretching&quot; criteria to cover the patient&#x27;s condition (or vice versa) is not at all uncommon. (Also, there&#x27;s not assigning diagnoses where it might negatively impact payment.)<p>EHR clinical data follows the discrete assumptions of the system design. This can create an impedance mismatch between clinician observations and data input. This may be troublesome, for example, in specialties (behavioral health) where data is complex with many overlapping subtle but meaningful variations.<p>I&#x27;ve often received records after a patient has been hospitalized. More than not it&#x27;s difficult to understand the course of treatment as there&#x27;s no coherent summary or narrative description provided. The &quot;pile of data&quot; literally transcribed from the EHR isn&#x27;t very useful to human readers. To be sure information like lab reports, etc., are good to have, but the marginalized human-to-human element is troublesome.<p>EHR clinical data failing to serve ML purposes could only point to problematic EHR system design. Though I&#x27;ve been aware of EHR limitations, I wouldn&#x27;t have guessed about ML issues. The article taught me something about the problem that exists. Now it remains to be seen what can or will be done about it.
btownabout 3 years ago
It strikes me that the decades of academic literature that establish &quot;in order to get a result correlating X and Y, we needed to control for A, B, C, ...&quot; would be a critical input into any system attempting to work with medical data. In a way, the plaintext of historical medical journals has encoded much of this expert knowledge, albeit with retractions and errors the further back you go. But that might change the conclusion of the OP into more of an &quot;incredibly hard problem&quot; rather than &quot;doomed to fail.&quot;
iancmceachernabout 3 years ago
You see these same pitfalls in much of the medical device industry. Often hospitals and their weird political internal workings drive reasons for things, and not quality or efficiency of care, care for their workers, etc. You see this present itself af far up as to help choose a certain technology development path for an entire industry based on these non-real internal hospital dynamics and intrenched ways of doing things, and billing for them rather than working to improve quality of care or outcomes.
tomlueabout 3 years ago
Spurious correlations are common in medical models not idiosyncratic like in other fields.<p>Which makes it hard to use models in clinics.<p>Doctors are smart about this. Every providier we spoke to about survival models based on biomarker data did not want supervised learning models. The explainability, trust, authorization just aren&#x27;t there and the risk of misuse is too high.<p>For now, research is a better use case, but less commercial funding.<p>It is not hopeless, but your basic LSTM is probably not going to revolutionize medicine.
apwheeleabout 3 years ago
I&#x27;ve always taken part of my job as a data scientist is to articulate when data is not sufficient to meet a particular a goal (and to outline what data would be necessary). For some goals doing special data sampling and building a model on small (complete) data is better than using the fragmented overall database.<p>I work mostly on claims data now, and building models to identify problematic claims I assure folks is feasible.
verisimiabout 3 years ago
What if I do need to go to a hospital, but don&#x27;t want any information about me to be shared? Is that impossible? What are the ethical considerations for me, when I disagree entirely to any data collection? Must I be coerced against my will into providing to Google or whoever regardless?<p>The point I am raising is that there is an embedded assumption that this info should even be available for deep learning to work over.
评论 #30768761 未加载
评论 #30767647 未加载
citizenpaulabout 3 years ago
You may not realize that EMRs owe their existence to<p><pre><code> 1.billing 2. government mandates 3. billing 4.helping doctors keep track of their patients’ records </code></pre> just like how people think ADP is in the business of payroll. Theyare actually in the business off mitigating regulation, taxes and liability. Getting your wage to you is at best 2nd priority to all those.
wistloabout 3 years ago
A rule encompassing &quot;pap smear&quot; and &quot;over 65&quot; data fields is encoded right here in the OP comment. Aren&#x27;t these kinds of rules and relationships what &quot;deep learning&quot; is supposed to suss out, automatically and without intervention?<p>If not, I wouldn&#x27;t call it &quot;deep&quot;.
naveen99about 3 years ago
If anyone is interested in working with emr data, my lab is looking for phd students, postdocs. We have quality anonymized emr and dicom data and research protocols to work with along with domain expertise and deep learning infrastructure.
myrryrabout 3 years ago
This seems like a lot of very &quot;USA&quot; problems.<p>A lot of countries don&#x27;t have the same drivers.
jmuganabout 3 years ago
We need to work on the hard problem of building causal models of the world, then we can build causal models of medicine on top, then we can do learning on medical records.
nikanjabout 3 years ago
But not before a lot of companies make tons of money by promising deep learning will cut healthcare costs by percentage points!
BurningFrogabout 3 years ago
This is very specific to the US medical system.<p>Perhaps things are better in some of the many other health care systems around the world.
评论 #30766758 未加载
评论 #30766646 未加载
mechanical_bearabout 3 years ago
These sort of headlines are just cheap click bait.
teleforceabout 3 years ago
It&#x27;s kind of funny that the author mentioned about data fragmentation in the very first section and at the end went to provide Heart rate&#x2F;ECG&#x2F;Afib as one of the successful case studies for deep learning for EMR.<p>ECG is the classic case of data fragmentation that he&#x27;s talking about and if you want to export raw data from the popular ECG machine like Mortara, good luck with that. Heck even Apple currently does not support exporting raw data ECG out of the box that&#x27;s crucial for deep learning, apart from the PDF image file for the ECG waveform [1]. Literally there are more than a few dozens open formats for the ECG [2],[3] and this obligatory XKCD comic comes to mind except that there are 39 standards instead of 15! [4]. Even if the ECG machine manufacturer is using one of the formats there are still serious interoperability issues down the road.<p>Rambling aside, there&#x27;s yearly cardiology global challenge competition organized by Computing and Cardiology Conference (CINC) and recently there are machine learning and deep learning techniques proposed for multi-lead ECG diagnostics but the results are not that great [5]. Hopefully it will start an impetus to deep learning in EMR similar to ImageNet Challenge that gave rise to the actual deep learning algorithm that drived the community past the winter AI.<p>[1] Accessing the ECG Data of the Apple Watch and Accomplishing Interoperability Through FHIR:<p><a href="https:&#x2F;&#x2F;pubmed.ncbi.nlm.nih.gov&#x2F;34042901&#x2F;" rel="nofollow">https:&#x2F;&#x2F;pubmed.ncbi.nlm.nih.gov&#x2F;34042901&#x2F;</a><p>[2] A review of ECG storage formats:<p><a href="https:&#x2F;&#x2F;pubmed.ncbi.nlm.nih.gov&#x2F;21775198&#x2F;" rel="nofollow">https:&#x2F;&#x2F;pubmed.ncbi.nlm.nih.gov&#x2F;21775198&#x2F;</a><p>[3] A Review on Digital ECG Formats and the Relationships Between Them:<p><a href="http:&#x2F;&#x2F;diec.unizar.es&#x2F;~imr&#x2F;personal&#x2F;docs&#x2F;paper12IEEETITB1.pdf" rel="nofollow">http:&#x2F;&#x2F;diec.unizar.es&#x2F;~imr&#x2F;personal&#x2F;docs&#x2F;paper12IEEETITB1.pd...</a><p>[4] How standards proliferate:<p><a href="https:&#x2F;&#x2F;xkcd.com&#x2F;927&#x2F;" rel="nofollow">https:&#x2F;&#x2F;xkcd.com&#x2F;927&#x2F;</a><p>[5]The PhysioNet&#x2F;CinC Challenge:<p><a href="https:&#x2F;&#x2F;cinc.org&#x2F;physionet-cinc-challenge-awards&#x2F;" rel="nofollow">https:&#x2F;&#x2F;cinc.org&#x2F;physionet-cinc-challenge-awards&#x2F;</a>