You'd have a blast around here...geography and city/county boundaries make our lives interesting. Let me explain...
We actually run in two systems at the same time: The Big O is our medical director in the City, and we have a couple of surgeons who we practice under in the county. So for the three years we had Amio before it was protocol, whenever we worked an arrest, we had to physically stop, figure out which side of the city/county line we were on, and then figure out which drugs we could push. It's the same now...we've got MgSO4 for severe asthmatics in the City now, as well as few other novelties, and it's funny whenever we run a City call and take the patients to the County MD's ED, the looks they give us are hilarious.
The county docs are the more conservative of the bunch, but even they are pretty damn liberal. The general rule is that protocols are only guidelines. I've been doing ACLS-level care for 7 years now, and I don't think I've ever had to call for permission to give a drug, and that includes using all the morphine we have, running dopamine and epi drips, etc.
I've exceeded protocol many times, but blatantly broken it only 3 or 4; on each occassion I fessed up to it BEFORE they asked, giving them my reasoning, and they've never had a problem with me.
At any rate, another county doc feels that Lido is still better than Amio, so he's going to start a new Lidocaine study. Problem is, they took lidocaine out of the box last year. So there's yet another drug we're going to have to lug around outside of the box (we had to carry amio that way for a long time).
Now I'm curious: do your medics have RSI? The county firefighters have it out here. My observation is that the results have been dismal. I once worked a day in the ED where they brought in two consecutive medical patients from a nursing home, both of whom they RSI'ed, and both of whom they ended up having to cric, and both of whom died. I should explain - their OMD, being another surgeon, insists that after two intubation attempts, they have to cric (!). I think the worst part is they're using Vec - if they'd sux 'em they could bag them till it wore off. I understand Vec for trauma patients, but not for medical. Personally I think RSI is a great technique, but without quite a bit of time spent tubing people in the OR, I think the Risk/Benefit ratio is way too high.
Fair enough. Amio has actually been in the US since the 60's. But we were the first to have it for Recurrent/Refractory VF/VT in the prehospital/ACLS setting, and we had it 3 or 4 years before it was even in ACLS.
The current ECC Guidelines 2000 are still wrong - because ACLS has always been epi-shock-drug-shock, Richard O. Cummins decided to set the algorithm as Epi-shock-Amio-Shock. Around here we give the Epi and the Amio togther - the Epi offsets the alpha and beta blocker action of the Amio. There's a study out (sorry, don't have a URL) that compares the two different algorithms - Epi with Amio and Amio after Epi....the combination at the same time is far superior to giving them sequentially.
For statistically based systems this is true, for neural networks it is not...all depends on what you use.
But you are right. There's a rule in ECG interpretation - when you hear hoofbeats, you expect to see a horse, not a zebra. In laymen's terms, it means that when you see a particular phenomenon, you expect it to be the thing you see 99.5% of the time, not the thing you see 0.5% of the time (in this case SVT with aberrant conduction).
Problem is, I worked on the electrophysiology floor, which, in equine terms, is where all of the zebras in the hospital were placed.
Dr. O spends more time developing new technology than practicing medicine (time well spent if you ask me), so he tends to know quite a bit about the innards of things. To put it succinctly...he's _well_ connected. In his defense: we had amiodarone for recurrent VF in 1997; we were the first agency in the country to have it. It's just now becoming a standard (that fact that the patent expired and the price dropped from $70 to $17 in November helped - again, I know because of O).
Field thrombolytics? Nah. I've talked to O about that. Only 1 out of 200 calls we run is an actual MI; to get the meds in every truck out here, we'd have to spend well over half a million dollars, just so that we could get orders to give TNK to one out of 200 people. Our Patient-contact to ED Door time here can be minimized to a max of 10 minutes (we're urban), and with Dr. O's typical 11-minute door-to-needle thrombo time, it's would be a monumental waste. The potential for rural thrombo is even more dismal, if you consider the volume of calls they run.
I'd like to thank you for taking the time to direct an EMS agency...it is quite noble of you.
Damn. In case you didn't know, many, many healthcare providers consider this book to be the bible of ECG interpretation, although I prefer this one.
The legalese page is missing from mine, but there is in fact a request to send interesting ECG's to an address in Florida, where the child porn charges went down.
I would have liked a 16.7% chance at that Thunderbird:(
Heh, it's nice to have an MD here on/. On a daily basis I use the Physio/Medtronic GE-based 12-lead systems in an ambulance, and then I have to put up with the recurringly Make-Me-Want-To-Go-Postal VF/VT alarms of HP/Philips Viridia systems in the hospital.
If I see "Nonspecific T-Wave abnormality" or "[Pick a location] Infarct, age indeterminate" on another ECG, I'm gonna tube someone's esophagus!!! As I said in an earlier post, The Viridia engineers are still using statistically-based FFT methods, which completely belies the process by which you and I interpret ECG's (although to be fair, the cones and rods in our eyes do perform an FFT). Also, their sample size is way too small (how many ECG's did you have to study before you knew what you were doing??? Was it more or less than 250??? I thought so).
At any rate I find the GE system in the Lifepak 12's far more accurate than the Viridia systems, though as I stated in an earlier post, every time I'm in the middle of diuresing a CHFer, the VT alarms start going off...and adjusting the amplitude of the ECG shuts them up.
Being a physician you might not understand the importance of the fact that if amplitude adjustment is a factor in the diagnosis, then from a purely signal-processing point of view, the diagnosis algorithm is complete crap. The LP12's are supposed to have the best noise filtering in the business (1), but I see wavy baselines on a daily basis...which from an engineering point of view is depressing.
Where engineers have failed, past the FFT model, is a lack of proper programming. If they would use fuzzy logic instead of statistics, then they would be able to translate linguistic rules into mathematical functions. For instance, fuzzy logic allows you to say "If the QRS is Wide (not specifying a particular time interval), and if there are broad, monomorphic R waves in I and V6, with no Q waves, and if there are broad, monomorphic S waves in V1, with possibly a small R wave, then it is LBBB." (Your interpretation criteria may vary). There is no analogue to this with statistics. You simply can't do it.
This is not to mention that going fuzzy would allow them to add info on troponins, CK-MB's, family history, previous MI's, etc. Fuzzy Logic is the way that YOU think when you are looking at an ECG. What's depressing is that not only have the failed to mimic human intelligence, they have failed to augment it. There are literally 3 or 4 hundred rules to evaluate when looking at a 12-lead. It's hard for a human to go through a list of a few hundred rules, but easy for a computer to do. Computers should be WAY ahead of cardiologists in this particular arena, yet they are far behind. And this goes for the paced rhythms and for the digitalis effects and for countless other things (and again, fuzzy rules let you specify that the patient is on dig or even digitoxic, which completely skews the interpretation).
I won't even mention that ECG's are chaotic, and there's not an interpretation system on the market that takes this into account.
Yes a human will still make the decision. But the computer is capable of so much more- predicting VF minutes before it happens, deciding WHEN to shock (an area of research during college for me) - that it's depressing to see physicians deciding to ignore the computer. Computers shouldn't be disagreeing with physicians - they should be augmenting their decisions, and giving them nonintuitive information.
--V-- (1) This is according to my medical director, Joseph P. Ornato (look in the front of your ACLS manual).
Yes you do. 97% of the 12-leads that I do spit out the diagnosis of "Nonspecific T-wave abnormality" and "Possible infarct, age undetermined". EVERY ECG has to be examined by a real-live person. which makes money spent on unreliable diagnosis software a complete waste.
BTW they are called EKGs. Yes, the letters are wrong, but it is to prevent people from confusing them with an EEG
Actually they are called ECG's. Many many Americans incorrectly use the term EKG, which is only correct if you are a Nazi, where it's spelled ElectroKardioGram. And since "Normal Sinus Rhythm" is not a phrase applied to EEG's, there's not the slightest chance of confusion.
And the interpretations still suck. Seems like every CHF patient I treat with wide-complex rhythm (IVCD, intraventricular conduction delay) ends up setting off the VF/VT alarm (now's a good time to look at my nick). The most embarrasing part about the VF/VT alarms is that if you change the amplitude of the ECG, the alarm shuts up....that should never happen!!!! VF and VT should _not_ be ruled out on the basis of a multiplication by 2:)
The problem is that to do the rhythm recognition, they're using a method of FFT-template matching. WHY?!?!?! Millions of healthcare workers interpret ECG's eveyrday without doing any FFT's. By using the template-only method, they are forgoing the method that humans use - morphology. So instead of looking at QRS width, P-wave dissociation, Brugada's and Josephson's signs, they look at an FFT. And they wonder why hospitals still have to employ ECG tech's to turn off all the false alarms.
Whether it be $4 or $10, it's a helluva lot cheaper than the roughly $38,000 that people around the country are paying for Lifepak 12 monitors. And I would know, being a paramedic. I actually had to do quite a bit of ECG processing during my stint at the local college to get a bachelors degree...I can't decide which is worse for ECG processing, Matlab (which I used) or Visual Basic. Matlab is nice, but at $3000 a pop, Visual Basic is the cheaper way to go. Matlab's built-in functions definitely helped tho (***cringe*** at the thought of implementing the FFT in VB).
Perhaps the biggest problem with ECG systems is noise filtering. Adaptive filtering has failed miserably; you'd think the engineers designing the systems would implement Bass and Treble-style filters for healthcare providers to use, but condescention has gotten the better of them. Noise makes interpreting an ECG sometimes impossible, yet no such filters. This presents a problem for me on a weekly basis - it is sometimes impossible to interpret the rhythm, because the filters aren't working properly. If they'd only let me adjust the lowpass and highpass cutoff's, I would feel SOOOO much better. And if they'd give me a Savitsky-Golay filter, I'd have a wet-dream....here's hoping.
Re:My wife the nurse said ...
on
Build Your Own ECG
·
· Score: 3, Informative
It's easy to say that it's possible to automate a lot of this stuff. It's harder to actually implement the automation. Philips owns Viridia this week, and their system is essentially the best on the market.
Unfortunately, the *best* in the market can only interpret QRS complexes. ECG techs still have to go into the system by hand, and label the PR, QRS, QT, and RR intervals. This is still quite an active area of research.
And for the record, they're nowhere NEAR 95% accurate. 70% would sound more accurate. And I would know, having been an ECG tech at a the local teaching hospital. The hospital spent $110,000 on the system and we were so unhappy with it that the engineering team came down from Massachusetts.
The rub? They're using a statistically-based FFT program, and their sample set is ~250 AHA ECG recordings. Humans who are good at ECG interpretation need to expose their neuro-fuzzy brains to at least 2,000 ECG's in order to know what they're doing. And they wondered why we were having problems.
When the head engineer got up to speak, he made an interesting comment. He said that when he was a student at MIT, a physician from Beth-Israel Deaconess Hospital came to the engineering department and asked if they could analyze ECG signals. They looked at them, and seeing their simplicity, said of course. Thirty years later, he's still working on the problem.
You'd think he would have tried something other than the FFT by now, but he hasn't. So much for thinking outside of the box.
I've mentioned this in a previous post, but just to make sure you understand: C# didn't evolve from Java. C# is Microsoft's way of monopolizing against Sun. It's not a derivative or descendant; It's a direct copy with enough "new" stuff so as to make it A) impossible for Sun to sue them and B) different enough to pass through the idiots at the USPTO.
1. Re-usability. Sun offers you a crapload of very usefull re-usable objects in their JDKs, and people like the Apache project offers you even more. The ability to do insanely complex prrojects with tiny amounts of effort is one reason why Java rules the corporate enterprise.
No, Java rules the corporate enterprise because Sun marketed it. The non-technical bosses in companies pushed their developers into using it. Its success is a function of buzzword popularity.
2) A universal computing environment: you can't write to the metal in Java, but it's not as slow as interpreted languages like Python and TCL for gigantic computing tasks. Any project, no matter how monolithic and task-optimized, is as portable as the VM is. Anyone who's had to manage a platform migration for key buisiness applications, from VAX to Solaris, say, or worse, from S/360 to Windows, knows the pain of re-implementation. That pain is gone when you use a VM-based language like Java.
That pain is also gone is Lisp. Java is 90% of Lisps portability. The difference is that in Lisp you move the source. If you need speed, you can compile to machine code without rewriting code.
This isn't to say that other languages aren't going to evolve, too, or are useless because they're not like Java. Ayone who programs in the new interpreted scripting languages: PHP, Python, Perl, Ruby, TCL, Scheme, etc, etc, can attest to the power of the that approach to modern computing.
Each of the languages you describe - PERL, Python, Ruby - are progressively closer and closer to Lisp. Scheme is, in fact, a dialect of Lisp. It's a good thing that you proved the authors point, since he undermined it. Also, the "new" things that make Java and other languages so popular - Garbage Collection, for instance - have present in Lisp since 1958. To quote Graham directly: "It's 2003, and we've almost caught up with 1958."
On the other hand, I really don't see any new compiled-to-the-metal languages emerging. Fortran is used for high-performance computing, Forth is used for tiny computers, and C/C++ is used for system programming. It will very likely be the same way in another 20 years, or another 50. The difference is that applications will slowly drift to either VM languages or interpreted languages from binaries compiled from source.
That's why he talks about code profilers - you write in a highly abstracted language, the profiler "talks to the metal" for you.
I think you are confusing "evolution" with an attempt by Microsoft to circumvent the fact that they lost the lawsuit against Sun. C# and.NET are not much more than attempt to nullify the need for Java. Why do you think M$ took Java out of XP? And look what happened - a judge made them put it back in. So to recap - evolution != monopoly (or for you lispers:
I had to switch to 0 threshold to see this and you AC'ed it in, but I thought I'd reply anyways.
On the one hand, you're quite correct - I should use another goddamn distro. Linux (and to be fair, FreeBSD which I also use) and OSS in general allow me to choose what I use, so I have no right to bitch about it.
Then there's the reality (and absurdity) of the situation.
I do a lot of programming, but that doesn't mean I should have to (nor do I want to) do a lot of system administration. I shouldn't have to add a terminal icon to my desktop or my main menu - the POWER of *nix is the command line. Does anybody think taking it off was a good idea?
Not only do I hate doing sysadmin tasks, when I'm not in a REP loop in Lisp, I just wanna be a simple user. So when RH switch from one WM to another without even mentioning it, or when they add (vertise) really excellent font support but fail to make obvious that you need to put your fonts in ~/.fonts, I get pretty pissy. But you know what? That's my right - every October and April I go to Best Buy and buy their software. So strictly as a user, I have a problem when the company I pay for software releases product that I pay for and then have trouble with.
If you don't like that, READ ANOTHER GODDAMN POST! If there are too many people bitching on/. for you, READ ANOTHER WEBSITE THEN!:)
I know I should calm down, and the terminal on the desktop isn't forbidden. But the sawfish thing turned out be be a real PITA. And like I said - upgrades should be easy, relaxing, and surprisingly impressive. 8.0 achieved the latter, but its sweetness was tainted with the bitter aftertaste of once again being unfamiliar with an OS that I had gotten comfortable with.
Redhat seem to have forgotten that many people won't use a *.0 release...now I've got to wait till October to upgrade my 7.3 box:(
Also, many commerical apps - for instance Franz's Allegro CL, which I use all day - won't support 9.0 for a while (they've just got around to supporting 8.0 this month).
Doing this to be at the same number as Slack??? Why not just switch to a sideways 8 and be done with that lame sort of one-upmanship already?(yeah, you're right - Volkerding would release "Slackware Infinity plus 2" - so then you'd release "Redhat Infinity plus Infinity").
Finally, a note to the RH people reading (bero-RH used to respond to all of my posts here on/. , but you guys kind of terminally pissed him off):
WTF is up with Metacity?
You put a window manager on your distro that doesn't even have a webpage? Many of us loyal RH users had gotten quite used to Sawfish. So used to it, in fact, that I had used GIMP to create many of my own themes.
So I was unpleasantly surprised, upon installing 8.0, to find that you guys had once again skipped a version number. What were you thinking? Didn't you you guys learn anything from the gcc-2.96 fiasco? (read the very bottom of the page). With 8.0 you've done it again - SF's sourceforge site has the most recent version at 1.2, yet somehow something named sawfish-2.0 made it into your distro. I frankly wouldn't care if your "2.0" worked; but whatever genius in NC decided to "upgrade" it forgot to also upgrade the sawfish-themer. The problem is that you also changed your entire font structure, so that SF2 barfed at my TTF bankgothic fonts. And then I had no themer to change it. Editing theme.jl by hand is a pain.
Now don't get me wrong. Metacity is a great WM. The fact that it uses XML is quite cool. And Havoc is a great programmer. But the fact that you switched WM's on us, and switched to essentially a wholly undocumented WM (there are pages now but none at release and really few for the first few months) is unacceptable.
Now I've kind of gone off on a tangent here, but I'm using a simple example to illustrate a very important point: Whomever is making the UI decisions at RH needs to stop it. I submit to you that some software companies do this thing called research - they find out what their users are using, and then make their products acceptable to them. Yeah, I know, I'm one of few people who customizes his own saw themes. So at this point I will also remind you that with 8.0 you took the terminal icon off of the desktop and the menubar and hid it 3 or 4 levels deep in the menu.
And a litany of other things - my point is that each release should have me sighing relaxing sighs of "Oh, this is nice," but instead every April and October I find myself feeling ever more uncomfortable and having to re-learn your entire distro. And don't tell me this is "innovation" - I know progress when I see it, and this isn't it.
On a better note, for the day or two that I did use 8.0, the fonts were beautiful - if you guys can hurry up and create an upgrade that I can live with, I'll love you for it.
The head of Økokrim's computer crimes division said that the reasons for the appeal would be in their offices Monday afternoon.
I guess they had a meeting with the MPAA today?
To the /. editors
on
Lego Segway
·
· Score: 4, Interesting
Could one of you lil' PERL monkies do us all a favor?
In the article submissions form, put a little check box titled "Slashdot can mirror locally" or some other phraseology.
Then provide the original link like you normally would, but on the last line where it says "Read More | XX of YY comments" add another link that says "Slashdot Article Mirror"
You seem to be unfamiliar the concept of proof by cases. A proof by cases is valid if and only if the cases are exhaustive.
You seem to be unfamiliar with the text of my post. Did I not say that for a theorem to be correct, that one has to prove its validity for ALL cases?
The proof of the Four Color Theorem broke the problem, or some lemma used in the problem, into around 1000 cases. The cases were exhaustive, or it would not have been a proof. Some curmudgeons didn't like the fact that the cases were checked by computer
I never said the first thing about the Four Color Theorem, only proofs in general. But I can tell you right now why these "curmudgeons" balked at it: one of the goals of mathematics is to express the... well, nature of nature as succinctly as possible. Which is better - having a long list of all of the cases that you've checked, or having a one-line equation that describes the system? (Hint: the root of this posting on/. was that Wolfram said he could express the Universe in 3-4 lines...think about it).
Would Schrodinger be a (geek's) household name if he had submitted a list of a thousand valid cases instead of his famous equation? Absoultely not - Heisenberg had already submitted a similar equation. The reason Schrodinger won fame was because his equation made it much easier to visualize the nature of particles, and much easier to solve problems.
is that the observable universe is defined by calculus and differential equations in very small areas: planetery motion, for example, or atomic physics.
No. The universe can be described by calculus and difficult equations.
Phenomena like life, geology and the like are very badly behaved with respect to our standard mathematical tools and we all know this.
No they're not. It's just that in order to model biological phenomena very well, you have to do finite state analysis on a very fine scale - and getting even the most powerful supercomputer to do the (calculus and difficult equations) calculations on ALL those elements is...unrealistic.
Still don't believe me? Ask IBM - according to their Blue Gene Project, given all the computing power available on the planet right now, it would still take several hundred years to calculate the structure of human proteins.
Still not convinced? OK. The human body is around a trillion cells. Each of these cells has millions of molecules in it. Each of those molecules can be accurately modeled by calculus and DifEq...but you've got trillions of millions of equations to worry about. There's not enough RAM on the planet to hold that matrix. Also consider that you also have trillions of millions of possible boundary conditions.
Wolfram is suggesting that cellular automata provide a simple framework for examining the phenomena outside of the "magic circle" of the calculus: i.e. most of life and the universe.
Wolfram is also supplying the code for the book in Mathematica. I submit to you that Wolfram is trying to sell more software. Which isn't to say I don't own his books, or that I don't have 4.1 minimized on my desktop right now. Because I do. But when someone comes out and says that his book is the be all and the end all of our scientific quandries, I am...skeptical.
You've hit a sort of soft spot here. I love the alternative ways of thinking - wavelets, fuzzy logic, neural networks, AI, and yes cellular automata. Each of these fields has been able to solve (or simplify) problems that have plagued researchers. But every time a researcher uses one of these methods to solve a problem, he or she starts evangelizing it, and hurts its credibility. It's a problem of HYPE - and there's HYPE written all over this book. He should have just published it, sent a copy to some respected scientists, and let THEM speak about it. I could care less when I see fervent hype on the MTV. But I get sick when I see it in scientific publications.
Of course, for a long time we've confused hard science with the application of calculus, which has effected what we consider "science" to be: if it is not an equation, we don't think it's scientific.
1> go talk to some biologists
Biologists are friggin crazy. I know this because I am one:)
2> get used to it: equations got us this far, but after this it may be increasingly about computation.
Consider, for example, the Four Color Theorem [wolfram.com] - the only existing proof of which requires a lot of computer power to grind through cases. Is it a valid proof? Probably - but not to the standards of mathematicians who grew up in the pre-computer age, to whom an exhaustively checked list of cases does not look like mathematics at all.
You seem to be unfamiliar with mathematical proofs. Grinding through many cases does not a valid proof make. In order to prove a theorem, you have to verify its validity for ALL cases, and in order to disprove a theorem, you only have to find one case where it is not valid. Just because you ran your theorem on a supercomputer for three months does not mean you have proved its validity for all cases. Example: You are trying to prove some theorem, and you use only positive integers. The supercomputer runs for a year and finds no holes in your theorem. Then your girlfriend comes over and enters -1, and your supercomputer barfs at you.
We'll see how Wolfram's work fares over time, but my bet is that it will fare Quite Well.
I HOPE it fairs well - I'd love to understand WHY the second law of thermodynamics works as well. But whether it fairs well or not, it would have faired much better without all of the hype and simply on the merits of the work itself.
YaST2, SuSE's Administration, is probably the biggest reason to switch. As far as apps go, linux apps (including the kernel) are pretty much the same - the main difference between any two distro's is their admin tools.
SuSE tends to come with more apps, and their default install is much better (in RH7.2, for some reason, tcpdump doesn't go on by default).
I use RH7.2, but the admin tools kinna suck. Networking in particular. I haven't installed S8 on one of my systems, but I tried YaST2 on a friends box and it's quite nice.
And oh yeah, SuSE never shipped a beta C compiler.
(Counting the seconds until bero-RH replies to this)
instead of the much more rational goal of forcing Microsoft to release APIs, file formats, network protocols, and other such information
Not even M$ understands Win32.
For the file formats, try wotsit.org (also - I have movies with junk at the beginning that mplayer can play but wmplayer 8 on XP will barf on - again, not even M$ understands file formats).
For the network protocols, read the freebsd source code:)
also at that site is the price list. At $642 for each of those Pentium II boards (not including RAM), I think I'll stick with buying "jumbo-mini" Beowulf nodes for the time being.
Slashdot Mirror
You'd have a blast around here...geography and city/county boundaries make our lives interesting. Let me explain...
We actually run in two systems at the same time: The Big O is our medical director in the City, and we have a couple of surgeons who we practice under in the county. So for the three years we had Amio before it was protocol, whenever we worked an arrest, we had to physically stop, figure out which side of the city/county line we were on, and then figure out which drugs we could push. It's the same now...we've got MgSO4 for severe asthmatics in the City now, as well as few other novelties, and it's funny whenever we run a City call and take the patients to the County MD's ED, the looks they give us are hilarious.
The county docs are the more conservative of the bunch, but even they are pretty damn liberal. The general rule is that protocols are only guidelines. I've been doing ACLS-level care for 7 years now, and I don't think I've ever had to call for permission to give a drug, and that includes using all the morphine we have, running dopamine and epi drips, etc.
I've exceeded protocol many times, but blatantly broken it only 3 or 4; on each occassion I fessed up to it BEFORE they asked, giving them my reasoning, and they've never had a problem with me.
At any rate, another county doc feels that Lido is still better than Amio, so he's going to start a new Lidocaine study. Problem is, they took lidocaine out of the box last year. So there's yet another drug we're going to have to lug around outside of the box (we had to carry amio that way for a long time).
Now I'm curious: do your medics have RSI? The county firefighters have it out here. My observation is that the results have been dismal. I once worked a day in the ED where they brought in two consecutive medical patients from a nursing home, both of whom they RSI'ed, and both of whom they ended up having to cric, and both of whom died. I should explain - their OMD, being another surgeon, insists that after two intubation attempts, they have to cric (!). I think the worst part is they're using Vec - if they'd sux 'em they could bag them till it wore off. I understand Vec for trauma patients, but not for medical. Personally I think RSI is a great technique, but without quite a bit of time spent tubing people in the OR, I think the Risk/Benefit ratio is way too high.
--V--
Fair enough. Amio has actually been in the US since the 60's. But we were the first to have it for Recurrent/Refractory VF/VT in the prehospital/ACLS setting, and we had it 3 or 4 years before it was even in ACLS.
The current ECC Guidelines 2000 are still wrong - because ACLS has always been epi-shock-drug-shock, Richard O. Cummins decided to set the algorithm as Epi-shock-Amio-Shock. Around here we give the Epi and the Amio togther - the Epi offsets the alpha and beta blocker action of the Amio. There's a study out (sorry, don't have a URL) that compares the two different algorithms - Epi with Amio and Amio after Epi....the combination at the same time is far superior to giving them sequentially.
--V--
Statistically, yes. With a neural network, No. The net would adapt - almost as if to say "What bunch of freaks do you have me analyzing?".
Every ECG gets reviewed anyways. So if a human has to look at every ECG, money spent on analysis software - accurate or not - is a waste of money.
If a company could produce a program that negated the need for human analysis, that would be worth the cost.
But even then it would be up to the FDA. Just look at the stupidity they've applied to pathology slides.
--V--
For statistically based systems this is true, for neural networks it is not...all depends on what you use.
But you are right. There's a rule in ECG interpretation - when you hear hoofbeats, you expect to see a horse, not a zebra. In laymen's terms, it means that when you see a particular phenomenon, you expect it to be the thing you see 99.5% of the time, not the thing you see 0.5% of the time (in this case SVT with aberrant conduction).
Problem is, I worked on the electrophysiology floor, which, in equine terms, is where all of the zebras in the hospital were placed.
--V--
Dr. O spends more time developing new technology than practicing medicine (time well spent if you ask me), so he tends to know quite a bit about the innards of things. To put it succinctly...he's _well_ connected. In his defense: we had amiodarone for recurrent VF in 1997; we were the first agency in the country to have it. It's just now becoming a standard (that fact that the patent expired and the price dropped from $70 to $17 in November helped - again, I know because of O).
Field thrombolytics? Nah. I've talked to O about that. Only 1 out of 200 calls we run is an actual MI; to get the meds in every truck out here, we'd have to spend well over half a million dollars, just so that we could get orders to give TNK to one out of 200 people. Our Patient-contact to ED Door time here can be minimized to a max of 10 minutes (we're urban), and with Dr. O's typical 11-minute door-to-needle thrombo time, it's would be a monumental waste. The potential for rural thrombo is even more dismal, if you consider the volume of calls they run.
I'd like to thank you for taking the time to direct an EMS agency...it is quite noble of you.
Here's a few links you might find interesting:
Dr. O's latest project
Screw ST elevation - here's color doppler radar for MI's (another Ornato project).
If your medics aren't using capnography, they should be. (This server is flogged at the moment).
--V--
PS - We're lucky here. Not only do we have Ornato, we've got Ellenbogen
PPS: It's bad enough that my truck buddies call me a geek...how'd you know???
Damn. In case you didn't know, many, many healthcare providers consider this book to be the bible of ECG interpretation, although I
:(
prefer this one.
The legalese page is missing from mine, but there is in fact a request to send interesting ECG's to an address in Florida, where the child porn charges went down.
I would have liked a 16.7% chance at that Thunderbird
Heh, it's nice to have an MD here on /. On a daily basis I use the Physio/Medtronic GE-based 12-lead systems in an ambulance, and then I have to put up with the recurringly Make-Me-Want-To-Go-Postal VF/VT alarms of HP/Philips Viridia systems in the hospital.
If I see "Nonspecific T-Wave abnormality" or "[Pick a location] Infarct, age indeterminate" on another ECG, I'm gonna tube someone's esophagus!!!
As I said in an earlier post, The Viridia engineers are still using statistically-based FFT methods, which completely belies the process by which you and I interpret ECG's (although to be fair, the cones and rods in our eyes do perform an FFT). Also, their sample size is way too small (how many ECG's did you have to study before you knew what you were doing??? Was it more or less than 250??? I thought so).
At any rate I find the GE system in the Lifepak 12's far more accurate than the Viridia systems, though as I stated in an earlier post, every time I'm in the middle of diuresing a CHFer, the VT alarms start going off...and adjusting the amplitude of the ECG shuts them up.
Being a physician you might not understand the importance of the fact that if amplitude adjustment is a factor in the diagnosis, then from a purely signal-processing point of view, the diagnosis algorithm is complete crap. The LP12's are supposed to have the best noise filtering in the business (1), but I see wavy baselines on a daily basis...which from an engineering point of view is depressing.
Where engineers have failed, past the FFT model, is a lack of proper programming. If they would use fuzzy logic instead of statistics, then they would be able to translate linguistic rules into mathematical functions. For instance, fuzzy logic allows you to say "If the QRS is Wide (not specifying a particular time interval), and if there are broad, monomorphic R waves in I and V6, with no Q waves, and if there are broad, monomorphic S waves in V1, with possibly a small R wave, then it is LBBB." (Your interpretation criteria may vary). There is no analogue to this with statistics. You simply can't do it.
This is not to mention that going fuzzy would allow them to add info on troponins, CK-MB's, family history, previous MI's, etc. Fuzzy Logic is the way that YOU think when you are looking at an ECG. What's depressing is that not only have the failed to mimic human intelligence, they have failed to augment it. There are literally 3 or 4 hundred rules to evaluate when looking at a 12-lead. It's hard for a human to go through a list of a few hundred rules, but easy for a computer to do. Computers should be WAY ahead of cardiologists in this particular arena, yet they are far behind. And this goes for the paced rhythms and for the digitalis effects and for countless other things (and again, fuzzy rules let you specify that the patient is on dig or even digitoxic, which completely skews the interpretation).
I won't even mention that ECG's are chaotic, and there's not an interpretation system on the market that takes this into account.
Yes a human will still make the decision. But the computer is capable of so much more- predicting VF minutes before it happens, deciding WHEN to shock (an area of research during college for me) - that it's depressing to see physicians deciding to ignore the computer. Computers shouldn't be disagreeing with physicians - they should be augmenting their decisions, and giving them nonintuitive information.
--V--
(1) This is according to my medical director, Joseph P. Ornato (look in the front of your ACLS manual).
Yes you do. 97% of the 12-leads that I do spit out the diagnosis of "Nonspecific T-wave abnormality" and "Possible infarct, age undetermined". EVERY ECG has to be examined by a real-live person. which makes money spent on unreliable diagnosis software a complete waste.
BTW they are called EKGs. Yes, the letters are wrong, but it is to prevent people from confusing them with an EEG
Actually they are called ECG's. Many many Americans incorrectly use the term EKG, which is only correct if you are a Nazi, where it's spelled ElectroKardioGram. And since "Normal Sinus Rhythm" is not a phrase applied to EEG's, there's not the slightest chance of confusion.
And the interpretations still suck. Seems like every CHF patient I treat with wide-complex rhythm (IVCD, intraventricular conduction delay) ends up setting off the VF/VT alarm (now's a good time to look at my nick). The most embarrasing part about the VF/VT alarms is that if you change the amplitude of the ECG, the alarm shuts up....that should never happen!!!! VF and VT should _not_ be ruled out on the basis of a multiplication by 2
The problem is that to do the rhythm recognition, they're using a method of FFT-template matching. WHY?!?!?! Millions of healthcare workers interpret ECG's eveyrday without doing any FFT's. By using the template-only method, they are forgoing the method that humans use - morphology. So instead of looking at QRS width, P-wave dissociation, Brugada's and Josephson's signs, they look at an FFT. And they wonder why hospitals still have to employ ECG tech's to turn off all the false alarms.
--V--
Whether it be $4 or $10, it's a helluva lot cheaper than the roughly $38,000 that people around the country are paying for Lifepak 12 monitors. And I would know, being a paramedic. I actually had to do quite a bit of ECG processing during my stint at the local college to get a bachelors degree...I can't decide which is worse for ECG processing, Matlab (which I used) or Visual Basic. Matlab is nice, but at $3000 a pop, Visual Basic is the cheaper way to go. Matlab's built-in functions definitely helped tho (***cringe*** at the thought of implementing the FFT in VB).
Perhaps the biggest problem with ECG systems is noise filtering. Adaptive filtering has failed miserably; you'd think the engineers designing the systems would implement Bass and Treble-style filters for healthcare providers to use, but condescention has gotten the better of them. Noise makes interpreting an ECG sometimes impossible, yet no such filters. This presents a problem for me on a weekly basis - it is sometimes impossible to interpret the rhythm, because the filters aren't working properly. If they'd only let me adjust the lowpass and highpass cutoff's, I would feel SOOOO much better. And if they'd give me a Savitsky-Golay filter, I'd have a wet-dream....here's hoping.
It's easy to say that it's possible to automate a lot of this stuff. It's harder to actually implement the automation. Philips owns Viridia this week, and their system is essentially the best on the market.
Unfortunately, the *best* in the market can only interpret QRS complexes. ECG techs still have to go into the system by hand, and label the PR, QRS, QT, and RR intervals. This is still quite an active area of research.
And for the record, they're nowhere NEAR 95% accurate. 70% would sound more accurate. And I would know, having been an ECG tech at a the local teaching hospital. The hospital spent $110,000 on the system and we were so unhappy with it that the engineering team came down from Massachusetts.
The rub? They're using a statistically-based FFT program, and their sample set is ~250 AHA ECG recordings. Humans who are good at ECG interpretation need to expose their neuro-fuzzy brains to at least 2,000 ECG's in order to know what they're doing. And they wondered why we were having problems.
When the head engineer got up to speak, he made an interesting comment. He said that when he was a student at MIT, a physician from Beth-Israel Deaconess Hospital came to the engineering department and asked if they could analyze ECG signals. They looked at them, and seeing their simplicity, said of course. Thirty years later, he's still working on the problem.
You'd think he would have tried something other than the FFT by now, but he hasn't. So much for thinking outside of the box.
--V--
I've mentioned this in a previous post, but just to make sure you understand: C# didn't evolve from Java. C# is Microsoft's way of monopolizing against Sun. It's not a derivative or descendant; It's a direct copy with enough "new" stuff so as to make it A) impossible for Sun to sue them and B) different enough to pass through the idiots at the USPTO.
1. Re-usability. Sun offers you a crapload of very usefull re-usable objects in their JDKs, and people like the Apache project offers you even more. The ability to do insanely complex prrojects with tiny amounts of effort is one reason why Java rules the corporate enterprise.
No, Java rules the corporate enterprise because Sun marketed it. The non-technical bosses in companies pushed their developers into using it. Its success is a function of buzzword popularity.
2) A universal computing environment: you can't write to the metal in Java, but it's not as slow as interpreted languages like Python and TCL for gigantic computing tasks. Any project, no matter how monolithic and task-optimized, is as portable as the VM is. Anyone who's had to manage a platform migration for key buisiness applications, from VAX to Solaris, say, or worse, from S/360 to Windows, knows the pain of re-implementation. That pain is gone when you use a VM-based language like Java.
That pain is also gone is Lisp. Java is 90% of Lisps portability. The difference is that in Lisp you move the source. If you need speed, you can compile to machine code without rewriting code.
This isn't to say that other languages aren't going to evolve, too, or are useless because they're not like Java. Ayone who programs in the new interpreted scripting languages: PHP, Python, Perl, Ruby, TCL, Scheme, etc, etc, can attest to the power of the that approach to modern computing.
Each of the languages you describe - PERL, Python, Ruby - are progressively closer and closer to Lisp. Scheme is, in fact, a dialect of Lisp. It's a good thing that you proved the authors point, since he undermined it. Also, the "new" things that make Java and other languages so popular - Garbage Collection, for instance - have present in Lisp since 1958. To quote Graham directly: "It's 2003, and we've almost caught up with 1958."
On the other hand, I really don't see any new compiled-to-the-metal languages emerging. Fortran is used for high-performance computing, Forth is used for tiny computers, and C/C++ is used for system programming. It will very likely be the same way in another 20 years, or another 50. The difference is that applications will slowly drift to either VM languages or interpreted languages from binaries compiled from source.
That's why he talks about code profilers - you write in a highly abstracted language, the profiler "talks to the metal" for you.
I think you are confusing "evolution" with an attempt by Microsoft to circumvent the fact that they lost the lawsuit against Sun. C# and .NET are not much more than attempt to nullify the need for Java. Why do you think M$ took Java out of XP? And look what happened - a judge made them put it back in. So to recap - evolution != monopoly (or for you lispers:
> (eq monopoly evolution)
NIL
I had to switch to 0 threshold to see this and you AC'ed it in, but I thought I'd reply anyways.
/. for you, READ ANOTHER WEBSITE THEN! :)
On the one hand, you're quite correct - I should use another goddamn distro. Linux (and to be fair, FreeBSD which I also use) and OSS in general allow me to choose what I use, so I have no right to bitch about it.
Then there's the reality (and absurdity) of the situation.
I do a lot of programming, but that doesn't mean I should have to (nor do I want to) do a lot of system administration. I shouldn't have to add a terminal icon to my desktop or my main menu - the POWER of *nix is the command line. Does anybody think taking it off was a good idea?
Not only do I hate doing sysadmin tasks, when I'm not in a REP loop in Lisp, I just wanna be a simple user. So when RH switch from one WM to another without even mentioning it, or when they add (vertise) really excellent font support but fail to make obvious that you need to put your fonts in ~/.fonts, I get pretty pissy. But you know what? That's my right - every October and April I go to Best Buy and buy their software. So strictly as a user, I have a problem when the company I pay for software releases product that I pay for and then have trouble with.
If you don't like that, READ ANOTHER GODDAMN POST! If there are too many people bitching on
I know I should calm down, and the terminal on the desktop isn't forbidden. But the sawfish thing turned out be be a real PITA. And like I said - upgrades should be easy, relaxing, and surprisingly impressive. 8.0 achieved the latter, but its sweetness was tainted with the bitter aftertaste of once again being unfamiliar with an OS that I had gotten comfortable with.
Redhat seem to have forgotten that many people won't use a *.0 release...now I've got to wait till October to upgrade my 7.3 box :(
/. , but you guys kind of terminally pissed him off):
Also, many commerical apps - for instance Franz's Allegro CL, which I use all day - won't support 9.0 for a while (they've just got around to supporting 8.0 this month).
Doing this to be at the same number as Slack??? Why not just switch to a sideways 8 and be done with that lame sort of one-upmanship already?(yeah, you're right - Volkerding would release "Slackware Infinity plus 2" - so then you'd release "Redhat Infinity plus Infinity").
Finally, a note to the RH people reading (bero-RH used to respond to all of my posts here on
WTF is up with Metacity?
You put a window manager on your distro that doesn't even have a webpage? Many of us loyal RH users had gotten quite used to Sawfish. So used to it, in fact, that I had used GIMP to create many of my own themes.
So I was unpleasantly surprised, upon installing 8.0, to find that you guys had once again skipped a version number. What were you thinking? Didn't you you guys learn anything from the gcc-2.96 fiasco? (read the very bottom of the page). With 8.0 you've done it again - SF's sourceforge site has the most recent version at 1.2, yet somehow something named sawfish-2.0 made it into your distro. I frankly wouldn't care if your "2.0" worked; but whatever genius in NC decided to "upgrade" it forgot to also upgrade the sawfish-themer. The problem is that you also changed your entire font structure, so that SF2 barfed at my TTF bankgothic fonts. And then I had no themer to change it. Editing theme.jl by hand is a pain.
Now don't get me wrong. Metacity is a great WM. The fact that it uses XML is quite cool. And Havoc is a great programmer. But the fact that you switched WM's on us, and switched to essentially a wholly undocumented WM (there are pages now but none at release and really few for the first few months) is unacceptable.
Now I've kind of gone off on a tangent here, but I'm using a simple example to illustrate a very important point: Whomever is making the UI decisions at RH needs to stop it. I submit to you that some software companies do this thing called research - they find out what their users are using, and then make their products acceptable to them. Yeah, I know, I'm one of few people who customizes his own saw themes. So at this point I will also remind you that with 8.0 you took the terminal icon off of the desktop and the menubar and hid it 3 or 4 levels deep in the menu.
And a litany of other things - my point is that each release should have me sighing relaxing sighs of "Oh, this is nice," but instead every April and October I find myself feeling ever more uncomfortable and having to re-learn your entire distro. And don't tell me this is "innovation" - I know progress when I see it, and this isn't it.
On a better note, for the day or two that I did use 8.0, the fonts were beautiful - if you guys can hurry up and create an upgrade that I can live with, I'll love you for it.
from the article:
The head of Økokrim's computer crimes division said that the reasons for the appeal would be in their offices Monday afternoon.
I guess they had a meeting with the MPAA today?
Could one of you lil' PERL monkies do us all a favor?
In the article submissions form, put a little check box titled "Slashdot can mirror locally" or some other phraseology.
Then provide the original link like you normally would, but on the last line where it says "Read More | XX of YY comments" add another link that says "Slashdot Article Mirror"
that religious Islamic fanatics are perhaps *more* moderate than the BSA.
:)
Maybe the new partnership with Egypt will help the BSA to tone down its image?
I think it's a sad day when I don't know which organization is more extreme
You seem to be unfamiliar the concept of proof by cases. A proof by cases is valid if and only if the cases are exhaustive.
/. was that Wolfram said he could express the Universe in 3-4 lines...think about it).
You seem to be unfamiliar with the text of my post. Did I not say that for a theorem to be correct, that one has to prove its validity for ALL cases?
The proof of the Four Color Theorem broke the problem, or some lemma used in the problem, into around 1000 cases. The cases were exhaustive, or it would not have been a proof. Some curmudgeons didn't like the fact that the cases were checked by computer
I never said the first thing about the Four Color Theorem, only proofs in general. But I can tell you right now why these "curmudgeons" balked at it: one of the goals of mathematics is to express the... well, nature of nature as succinctly as possible. Which is better - having a long list of all of the cases that you've checked, or having a one-line equation that describes the system? (Hint: the root of this posting on
Would Schrodinger be a (geek's) household name if he had submitted a list of a thousand valid cases instead of his famous equation? Absoultely not - Heisenberg had already submitted a similar equation. The reason Schrodinger won fame was because his equation made it much easier to visualize the nature of particles, and much easier to solve problems.
is that the observable universe is defined by calculus and differential equations in very small areas: planetery motion, for example, or atomic physics.
:)
No. The universe can be described by calculus and difficult equations.
Phenomena like life, geology and the like are very badly behaved with respect to our standard mathematical tools and we all know this.
No they're not. It's just that in order to model biological phenomena very well, you have to do finite state analysis on a very fine scale - and getting even the most powerful supercomputer to do the (calculus and difficult equations) calculations on ALL those elements is...unrealistic.
Still don't believe me? Ask IBM - according to their Blue Gene Project, given all the computing power available on the planet right now, it would still take several hundred years to calculate the structure of human proteins.
Still not convinced? OK. The human body is around a trillion cells. Each of these cells has millions of molecules in it. Each of those molecules can be accurately modeled by calculus and DifEq...but you've got trillions of millions of equations to worry about. There's not enough RAM on the planet to hold that matrix. Also consider that you also have trillions of millions of possible boundary conditions.
Wolfram is suggesting that cellular automata provide a simple framework for examining the phenomena outside of the "magic circle" of the calculus: i.e. most of life and the universe.
Wolfram is also supplying the code for the book in Mathematica. I submit to you that Wolfram is trying to sell more software. Which isn't to say I don't own his books, or that I don't have 4.1 minimized on my desktop right now. Because I do. But when someone comes out and says that his book is the be all and the end all of our scientific quandries, I am...skeptical.
You've hit a sort of soft spot here. I love the alternative ways of thinking - wavelets, fuzzy logic, neural networks, AI, and yes cellular automata. Each of these fields has been able to solve (or simplify) problems that have plagued researchers. But every time a researcher uses one of these methods to solve a problem, he or she starts evangelizing it, and hurts its credibility. It's a problem of HYPE - and there's HYPE written all over this book. He should have just published it, sent a copy to some respected scientists, and let THEM speak about it. I could care less when I see fervent hype on the MTV. But I get sick when I see it in scientific publications.
Of course, for a long time we've confused hard science with the application of calculus, which has effected what we consider "science" to be: if it is not an equation, we don't think it's scientific.
1> go talk to some biologists
Biologists are friggin crazy. I know this because I am one
2> get used to it: equations got us this far, but after this it may be increasingly about computation.
Consider, for example, the Four Color Theorem [wolfram.com] - the only existing proof of which requires a lot of computer power to grind through cases. Is it a valid proof? Probably - but not to the standards of mathematicians who grew up in the pre-computer age, to whom an exhaustively checked list of cases does not look like mathematics at all.
You seem to be unfamiliar with mathematical proofs. Grinding through many cases does not a valid proof make. In order to prove a theorem, you have to verify its validity for ALL cases, and in order to disprove a theorem, you only have to find one case where it is not valid. Just because you ran your theorem on a supercomputer for three months does not mean you have proved its validity for all cases. Example: You are trying to prove some theorem, and you use only positive integers. The supercomputer runs for a year and finds no holes in your theorem. Then your girlfriend comes over and enters -1, and your supercomputer barfs at you.
We'll see how Wolfram's work fares over time, but my bet is that it will fare Quite Well.
I HOPE it fairs well - I'd love to understand WHY the second law of thermodynamics works as well. But whether it fairs well or not, it would have faired much better without all of the hype and simply on the merits of the work itself.
YaST2, SuSE's Administration, is probably the biggest reason to switch. As far as apps go, linux apps (including the kernel) are pretty much the same - the main difference between any
two distro's is their admin tools.
SuSE tends to come with more apps, and their default install is much better (in RH7.2, for some reason, tcpdump doesn't go on by default).
I use RH7.2, but the admin tools kinna suck. Networking in particular. I haven't installed S8 on one of my systems, but I tried YaST2 on a friends box and it's quite nice.
And oh yeah, SuSE never shipped a beta C compiler.
(Counting the seconds until bero-RH replies to this)
instead of the much more rational goal of forcing Microsoft to release APIs, file formats, network protocols, and other such information
:)
Not even M$ understands Win32.
For the file formats, try wotsit.org (also - I have movies with junk at the beginning that mplayer can play but wmplayer 8 on XP will barf on - again, not even M$ understands file formats).
For the network protocols, read the freebsd source code
also at that site is the price list. At $642 for each of those Pentium II boards (not including RAM), I think I'll stick with buying "jumbo-mini" Beowulf nodes for the time being.