Slashdot Mirror
Excerpt from Kessler's 'The End of Medicine'
With my head spinning from this "High Anxiety" flashback, I stroll into the conference, half expecting to be given a barium enema by a cross between Nurse Diesel from Mel Brooks' flick and Nurse Ratched from One Flew Over The Cuckoo's Nest. I really gotta switch to decaf on days like this.
The 7th International Multi-Detector Row Computed Tomography Symposium sounded innocuous enough. I assumed it would be a bunch of technical papers on the future of scanning, where I would read the paper in the darkened hall until lunchtime and then head off for some hot Hunan and home.
Instead, the place was like a carnival for cardiologists. Talk about feeling like a fish out of water. Outside the hall was an expo of sorts, with big signs flashing Toshiba and Philips. Instead of TVs or microwave ovens, there were PCs with 3D models of some poor schmuck's diseased coronary arteries being folded, stapled and rotated.
The back wall of Toshiba's booth caught my eye and I just stared at it. Rule number one at any tradeshow booth is never look interested or you are doomed to a rapid-fire ten-minute lecture on the ins and outs of the product and forced to give up your card as a qualified lead, to be hounded by phone, fax, email and snail mail for the next year.
"Those are our detectors." Damn, I was snagged.
"They look like the display on my laptop," I noted.
"Well sure, they are not that much different from a flat-panel display."
"Same economics making them?" I asked. Flat panels are notoriously expensive to manufacture, because of their size, unlike chips, where hundreds can fit on an eight-inch diameter wafer.
"Oh no, as we go from 4- to 16- to 64-slice, the detectors can be manufactured discreetly and butted up against each other. We don't have yield issues."
"How much is one of these 64-slice scanners?" I asked.
"Are you ready to buy one today, or this month?" booth-guy asked me.
"No, no, although I wouldn't mind one in my garage. I'm a tech guy."
"Oh, OK. Well, these are basically one- or two-million dollar machines."
"Wow." I wasn't sure if that is a lot or a little, but often a well-placed 'wow' gets you all sorts of inside scoop.
"I know, pretty cheap. We think we have a variety of advantages over the competition and you will see in the face-off that ..."
"Why so much? I've been in enough factories, and those flat panels are a couple of hundred bucks each and the motor to rotate can't be more than ..."
"Well, the X-Ray source is not inexpensive."
"What? Hundreds of thousands of dollars?" I trolled.
"Probably not. We do have high selling expenses. When you only sell a hundred of anything, there is lead generation and a sales pipeline and funnel."
He started whispering. "They could be a lot cheaper." He must be having a tough month.
"Don't let me stop you, by the way," I said, looking around, trying to imply he should hard sell some of these cardiologists and radiologists who were buzzing around the display.
"Doctors aren't buyers, not for these machines. We sell to a few clinics. The rest is into hospitals - they are the only ones that can afford them for now."
"But you said cheaper — I mean, these can be in the hundreds of thousands of dollars instead of millions." It was a statement dressed up as a question.
"Someday," he whispered, again.
That's all I needed to know.
Several times, I heard references to the big face-off that afternoon, like it was the reason everyone was there. "Don't miss the face-off," "This ought to show well at the face-off," "This year is going to be so much better than last year's face-off." OK, I get it.
I sat down in the auditorium and the talks and dim lights put me right to sleep.
The head whips woke me up, as my neck turned into Jello and my chin dug into my chest. I wasn't sure if I was awake, my heart was beating fast - I was on the top floor looking over the rail next to Mel Brooks ... Nope, I'm OK, I'm awake, although embarrassed as quite a few radiologists turn to see what the commotion was in my seat.
"Ladies and gentlemen, welcome back, take your seats, fasten your seatbelts, this is going to be exciting. I am pleased to announce that for our 3rd Annual Workstation Face-off, we have five different vendor groups competing — well, facing off. We have five different data sets: brain, runoff, lung, colon and heart."
The room exploded in applause, like this was some sort of important revelation.
"On the stage, we have workstations from GE Healthcare. Dr. Gruden, please take a bow. Also Vital Images, Philips Medical Systems, Siemens Medical Solutions and TeraRecon. May the best workstation win. Let's get started."
The room was buzzing. On stage were two giant screens. On the left was a view from the monitor of the workstation and on the right was a live feed from the operator's keyboard and mouse so the audience could see how many clicks and keystrokes and other contortions are needed to get through the data set.
"OK, let's start with the brain. GE, you have six minutes for both the Angiogram and the Perfusion. Go."
A giant clock on stage started counting down from 6 minutes. The doctor operating the GE workstation was furiously clicking and slapping his mouse around and on screen; we all could share his view zooming through someone's brain.
"OK, we can see the internal carotid artery on the right-hand side, so now let's quickly move over to this area on the left, ah, not hard to find, there it is, we see the ICA stenosis, let's measure it, 63% blockage." A smattering of applause. "We can zoom in and clearly delineate the calcified vs. the soft plaque." More applause.
"OK, let's quantify the infarct core ..."
I was transfixed. This guy was zooming through someone's brain like it was a Sunday drive. More like a Sunday afternoon video game. I kept looking for a brain in a jar of formaldehyde labeled "Dysfunctio Cerebri — Abnormal Brain" and Dr. Frankenstein's assistant Fritz limping back to the laboratory.
"Let's mark this tissue at risk for infarction and measure some things while we are over in the left cerebral - OK - MTT is 86.7, TTP let's call it 52.5 ..."
He zoomed around the brain like it was just a bunch of bits on the screen, which of course it was. Duh.
"OK. Time. 5 minutes 32 seconds. Very nice. Thank you," the moderator said. The place went crazy. This was repeated on each of the workstations by different doctors to often-thunderous applause. I had a mild headache from all the excitement.
I watched these workstations find aneurysms in the arteries from the waist down, the run off. The trick is to remove the bones from the view and be left with just the arteries. Jeez, everyone knows that. Even I could find the mild aneurismal dilation of right renal arterial trifurcation! But my feet started to hurt and I looked around and lots of folks were rubbing their calves.
In the lung, the fly-throughs were looking for lobe nodules, which weren't so obvious. It was a maze of tubes in there — who can even find their way, let alone in under 4 minutes? But sure enough, there was the posterior and the one adjacent to the heart. Each of the five operators then went back and compared them to a study from three years earlier, after finding them in the previous study, of course. Pretty cool. Does my doctor have this? I coughed, more of an unconscious reflex than anything else.
"OK, a perennial favorite, let's move on to the bowels. This year's virtual colonoscopy will require identifying and measuring five different polyps as well as comparing supine and prone data sets to differentiate stool from polyps."
There was a gasp from the crowd, probably from all the men over 50 who have not-so-fond memories of their real colonoscopies.
"The folks from TeraRecon will go first." "Thank you. For this data set, we have decided to show off our handheld interface device. It is a two-handed device, requiring minimal keyboard usage."
On the right-hand screen, the view zoomed into the doctor's hands wrapped around what looked like a Nintendo or Sony Playstation controller. He was banging it and twisting it around, not much different than my kids playing Halo 2. Except that on the left-hand screen, instead of you as Master Chief blowing away the Covenant to stop them from destroying Earth, you are Master Doctor searching for cancerous polyps extracting revenge and trying to destroy your patient. Or something like that. And you only have six minutes and a crowd of a thousand to cheer you one.
"OK," the doctor running the TeraRecon station said, "let's go into C.A.D. mode to navigate through the colon."
On screen, the screen started flying through the wrinkled walls of the colon, twisting and turning, to the left, sliding over, turning up, then right, around a corner, then down again until it saw something abnormal and stopped in front of a hanging polyp. Ah, that's what Steve Sandy was telling me about.
Massive applause.
TeraRecon found all the polyps and so did everyone else. It wasn't hard, those polyps hung like fruits from a tree, pretty obvious against the background of the empty colon. Each of the operators had to go to the alternate data set to show that a few potential polyp looking globes were nothing more than a pile of, well, stool.
My cough had mysteriously turned into a pain in my lower gut.
"Now, what you have all been waiting for, the grand finale, someone left their heart in San Francisco."
On screen was a giant rendering of a heart and most of the coronary arteries. It might as well have been pumping and spraying blood all over the audience like the movie Carrie, there was such a frenzy.
Each of the workstations zoomed in, probed for diameters of sinotubular junctions and aneurismal sinuses. Ho hum. But in no time, each found blockages, stenosis that either had already caused a heart attack or was about to any day.
I just stared at the screen. My eyes were wider than Marty Feldman as Igor in Young Frankenstein. It's not some dream of the future, there it is in front of my face. I felt some pains on the left side of my chest, but my stomach ache went away.
This is it. The resolution was high enough, and there was plenty of speed to zoom around and find all the gunk in less than five minutes. These guys could peak inside and tell me if I was going to have a heart attack, before I do, before I drop on the floor grabbing my chest and my wife screams to the 911 operator to get someone there as fast as they can, before all my relatives get the call saying Andy has had a heart attack, before I get overloaded with blood thinners and can't remember what day it is.
This changes everything. Blood pressure readings, cholesterol checks for low-density lipoproteins, echocardiograms, all that stuff is primitive stuff, like silent movies — OK, another Mel Brooks reference. It just has to be cheap enough and it will be as routine as the doctor banging your knee or squeezing the crowned jewels.
Let's see: $2 million machine, 5 minutes per patient, of course, that means
144 a day, 720 per week, 36,000 per year, hmmm, that's $55 per scan.
Add a little for the attendees and five minutes of the radiologists
time and voila, maybe this is a mass market thing after all.
Avoid any surgeon run by Windows, it might BSoD while working on your heart!
/. is not to be used by individuals with high blood pressure or a history of heart attacks
my computer doctor isn't plagued by 503 errors in the middle of surgery!
Monstar L
Spoke with one of my esteemed colleagues re this. There are still certain things that the CT virtual scan is not good at detecting - Not all polyps are pedunculated (like a tree) - some are broad and flat (sessile), some vascular lesions can not be appreciated with the CT, etc.
So what to do if you find a polyp? - get an actual colonoscopy of course, so that they can snip it off.
While most people don't look foward to having this done, it's still probably the best way to have your colon checked out. Everyone over 50 NEEDS to have this done. Missed colon cancer can lead to a colostomy - yeeecch! - or worse. So put up with the distastefulness of it and get it done, or convince your parents to get theirs.
..........FULL STOP.
As an undergrad in the 80s, I worked with some computers in a university chemistry lab. In this lab, one of the research professors was developing "shape fitting" methods to design drug molecules. Need to attack a certain receptor? Design a drug that fits. Need to protect a certain receptor? Design a drug that plugs the hole until the intended natural molecule is present. It was all very next-century super-futuristic stuff.
Now that computers should be able to handle that task easily, I rarely hear anything about it anymore. And honestly, there's a lot more than geometry and a few chemical bonds that need to be better understood. We all thought buckyballs would be completely inert and pass through the system... until we actually found that living bodies can get choked up with them. It's like explaining how bees fly-- there's a lot to the science which is still just guesswork and lab experiments.
Lastly, it may be great if a new treatment helps 94% of patients... unless you fall in the 6% it doesn't help. Everything is statistical until it's personal. There are a lot of areas of the response tree which are not known, or even if a certain branch of pathy is known, there are risks in all modes of treatment.
Somehow, I don't think human fuzzy-thinking seat-of-the-pants gut-instinct doctors will be replaced with deterministic analytical machinery anytime soon.
[
Finally, went to a third doctor who gave me a different Lyme test which came back borderline (but still technically negative). She put me on antibiotics for a few months, and thanks to that treatment, I'm much better (not as well as before, but about 95%) now. It takes a good diagnostician to listen to the patient's symptoms, ask questions about his/her history, and *not* blindly look at test results.
I'm not saying that this equipment isn't important, just that there's still a place for talented physicians - those things are an adjunct, not a panacea.
-b.
Mind you, I happen to have had an echocardiogram just last week, my first, and it's a freakin' miracle to see all the little valves doing their stuff, and a technician watching my heart in real time for many minutes and making literally dozens of quantitative measurements without poking sharp things into me or injecting dubious "dyes" into me or (I trust!) toasting me with radiation.
But I have to wonder. If high-tech medicine is actually effective--not just awe-inspiring, exciting, and, well, entertaining--why is it that with so much of the stuff, the United States ranks about #40 in infant mortality (worse than New Zealand, Portugal, Slovenia)? Why is our life expectancy only 78 years when forty-seven other countries, including Aruba, Spain, and Iceland, do better?
Is it possible that we need less of these robotic surgeons and computer imaging centers and a few more humble, prosaic things... like visiting nurses, or immunization programs (How is it possible that people in the United States are still getting mumps)?
"How to Do Nothing," kids activities, back in print!
One of the main theses of this book is the comptuer technology will do to doctors what ATMs did to tellers. I call BS. My wife is studying to be a doctor (MD), and its more than memorising disease X and treatment Y. It involves alot of bedside manner and gut instinct. Think about what a teller does... a tellers does not need 4 years of teller school and 3+ years of teller residency to do his or her job. A teller fscks up a transaction, no one's going to die. An doctor fscks a diagnosis, well the patient may be up the proverbal polluted creek without a means of propulsion (i.e. death).
I've done research into using AI methods to diagnosis patients with a disease based on MALDI-TOF of proteins from patients with and without lung tumors. The group I worked in had a difficult job spliting groups. When we presented at a conference, everyother presenter could not find and answer (this was a sponsored 'contest' to see if it could be done). It was a b*tch to seperate patients by biological markers. AI will probably be able to do it one day, but not now.
(paper: Proteomics. 2003 Sep;3(9):1704-9. Multiple approaches to data-mining of proteomic data based on statistical and pattern classification methods. Tatay JW, Feng X, Sobczak N, Jiang H, Chen CF, Kirova R, Struble C, Wang NJ, Tonellato PJ.)
Just my 4 bits...
Accentuate the positive, don't waste your mod points on the negative.
You get to a point with investigations where more isn't necessarily better. There is always a chance that a scan will show a false positive. You see what looks like a suspicious nodule in an asymptomatic patient. Now you have to do more investigations, and some investigations carry a risk. Finally they decide to biopsy the nodule to see what it is and it turns out to be benign. The end result can be a patient exposed to unnecessary radiation, surgeries and/or worries. That's why it's nice to have doctors who decide who and when to scan. 'Magic Scans' aren't the panacea they are cracked up to be.
That being said, it is incredible what technologies are doing for medicine. Some of the stuff in medical textbooks is obsolete, particularly clinical findings, because you almost never see that disease advanced to the degree that the book describes as we get better and better at finding things early.
So like the other poster said, sign up to get your screening colonoscopy today! (provided that your 50)
I have been running a radiology IT company for the last 4 years and have been involved with radiology in one way or another half my life (all of my life if you count my father as being a radiologist). I frequent medical technology trade shows and have seen quite a bit of this 3D post processing.. Very intriguing stuff and the resolution on these machines is very beautiful, I often wish my home PC could render the same resolution as these machines. Doom 6 would be the next best thing to sex..
However, I don't think any machine will be capable of replacing a medical professional in the next 20 years and I really feel there is only one major reason holding vendors back: liability. The human body is simply too complex for a computer to check for issues, understand complications, diagnose and recommend further action.
Currently the systems can highlight and pinpoint potential problem areas, but these systems still rely on a medical professional to review and approve the findings. From what I here the systems are pretty good at finding "problem" areas but offer a lot of false-positives that the doctor has to check out.. thus negating any upper hand advantage they gave in work flow optimization.
My brother currently has about 2-3 months to live because he didn't and for the last five years has battled cancer. Five bouts of chemo. Not something to wish on anyone. Please get that checkup.
In my experience algorithms are rarely monopolized by any single field very long and fairly quickly find themselves distributed across all sciences and engineering. For example the algorithm of tomographic inversion was picked up by seismologists, astrophysicists, meteorologists, material scientists, etc. for similar situtations in their fields. Likewise radiology engineers monitor devlopments in image processing and 3D graphics to construct even move vivid and useful body images.
It's pretty clear that better Lyme Disease tests would have been easier and more effective. Bring on the technology!
As an undergrad in the 80s, I worked with some computers in a university chemistry lab. In this lab, one of the research professors was developing "shape fitting" methods to design drug molecules. Need to attack a certain receptor? Design a drug that fits. Need to protect a certain receptor? Design a drug that plugs the hole until the intended natural molecule is present. It was all very next-century super-futuristic stuff.
Now that computers should be able to handle that task easily, I rarely hear anything about it anymore...
You have to find the receptor. You have to design a drug to fit in the receptor. You have to figure out (mostly guess) whether or not the drug will actually bind at the receptor. You have to figure out (at least guess) as to whether or not the drug will kill people in addition to stopping acne. You have to figure out how to make the drug. Then you have to actually go through all the trials.
If you mass test random potential compounds against random proteins, you get to cut out some of those steps.
There are companies that do the designer route, but it seems like most go for the shotgun "spray a bunch of compounds at a bunch of proteins and see what sticks".
Really, we don't even know how these proteins are gonna fold, so we're a ways away from automating designer drugs.
*I am not a biologist, chemist, physicist, etc. I'm a programmer.*
I was talking about the communicators in The Original Series, not from The Next Generation. They looked very much like today's flip phones.
Nor was I talking about stun guns. We do have honest to goodness laser weapons now, which at this point only cause blindness, but there are also weapons under development that will do further damage.
http://en.wikipedia.org/wiki/Personnel_Halting_an
There is also a laser-equipped 747 that can shoot high-powered beams at a given target.
http://www.defensetech.org/archives/000551.html
The tricorder is also a real device, although the technology still has a long way to go.
http://www.stim.com/Stim-x/0996September/Sparky/t
Transparent Aluminum, first introduced with ST4:The Voyage Home as the superstrong "metal glass" used to haul two humpback whales back to the 23rd century, is also the real thing.
http://en.wikipedia.org/wiki/Transparent_aluminum
I don't smoke anything, not even tobacco.
The problem with socialism is that they always run out of other people's money. - Margaret Thatcher
However, I googled him recently out of curiosity to see what had happened to him and his ideas, 20+ years on. I have to say that the results were remarkably disappointing. It seems that Larry Weed was too far ahead of his time and may still be. Check out his company at http://www.pkc.com./ One rather dramatic quote of his: if physicians ran airports they wouldn't have radar, just lots of intensive care units around the periphery.
(a) antibiotics don't treat Lyme immediately. It can take weeks to months to completely remove the bacteria from your body. Thus, the usual course of 10-days of antibiotics at a normal dosage would have done little or nothing (also, when you start antibiotics, Lyme often gets worse, not better).
(b) the Lyme symptoms are close to many autoimmune disorders, and thus can be mistaken for such. Thus, some doctors have prescribed steroids - corticosteroids lower immune responce, which is the exact opposite effect than the one desired.
-b.
It's like explaining how bees fly-- there's a lot to the science which is still just guesswork and lab experiments.
The problem with explaining how bumblebees fly was that the wrong model was applied. They tried to analyze a bumblebee as if it were a fixed-wing glider (or a helicopter - which is mainly a fixed wing aircraft flying in cricles) and discovered that it would drop like a rock.
Which it will if it stops flapping.
Took a while to figure out that they create a vortex with one part of the flap, then use it to create lift with another. But they've pretty much got it down solid at this point - and are building tiny flapping flying machines using the same principle that perform about as analysis predicts.
Of course the same is no doubt happening all over medicine. So your analogy is right on (even if the example is a bit dated).
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Speaking as a practicing physician (pediatric hospitalist, to be precise), there are at least three things that are going to keep me from worrying too much about being "rebooted" by a really good CT scanner.
First, there's the unpleasant reality than in medicine the diagnosis is usually not the most important question. I've had about three true diagnostic conundrums in the past two years, and in two of those the question wasn't what was wrong (we were virtually sure it was cancer) but where it was. Yes, in those two cases, a very high-tech scan (a PET/CT) helped make the diagnosis, but for every one of the other hundreds of patients I've seen recently, the key issue was management, where all the 3-d algorithms in the world are brutally inferior to one reasonably well-educated intern.
This is even more important because diagnosis isn't enough, ever. You can't just find the polyps, you have to deal with them. You can't just find the coronary stenosis, you have to repair it. And in both cases, the skill of the physician (and the knowledge to accurately measure benefit and risk) are my real stock in trade. Hand me the diagnosis, and I'm not threatened, I'm thrilled. And since you need a good radiologist to really read the CT well, my radiology colleagues are pretty OK with the new tech as well.
Second, the excerpt is coming from a radiology trade show. I'm glad that the tech is cool, and I love the pictures, but radiology is only a small part of medicine, and most of the non-trauma diagnoses we see aren't really that dependent on a good CT. As other posters have noted, echocardiography and endoscopy do pretty well at all of this, and the CT is at best a screening adjunct that might increase the numbers of people who have the definitive studies. Eventually the imaging will be good enough to really replace colonoscopy (just as it became the standard of care in diagnosis of appendicitis), and I really do hope that it happens before I get to 50. But it's a tiny part of medical care, blown into high relief because it's at a show where nothing else is important. It's like going to an embedded-systems trade show and not noticing that graphics exist, because nobody is embedding 3D in their network storage appliances.
And third, and possibly just because I'm being cynical, I've been replaced by various kinds of high technology since before I even became a doctor. I've been outmoded by fuzzy logic systems, by automated diagnostic software, by genomics, by proteomics, by targeted drug design, and by about fifteen different funding agency mandates. I've been told I'm obsolete so long that the first ones that told me are already dead. Sure, I expect to die myself some day - as far as I know, even in this age of high tech, everybody pretty much does die - but I'm not going to spend much time worrying that technology will make physicians obsolete before then. The game changes, and I practice medicine very differently from the way it was done in my grandfather's day (and thank God that this is so), but as long as I'm willing to employ my intellect and manual skills on behalf of sick kids, there'll be a way to do it.
Now that computers should be able to handle that task easily, I rarely hear anything about it anymore.
Nooo, I'd say you don't hear so much about it anymore because (1) the idea is no longer new, and (2) it isn't working out as well as we'd first hoped, and perhaps (3) you're not in the field. We've a loooong way to go before rational drug design in silico becomes truly routine. At the moment it's a big help to the trained chemist, but that's it.
A couple of problems remain:
(1) Often enough, the molecular structure of drug targets are a mystery. Most often, it's a mystery clinically, in the sense that no one knows the target molecule or even the target biochemical pathway. How do you fight the development of atherosclerotic plaques in the coronary arteries? Given that we have incomplete knowledge of how it happens, it's very hard to identify a target for drug therapy. Here genomics and proteomics, e.g. the correlation of the expression of certain proteins with certain clinical conditions will undoubtably help.
Even in cases where a target is known, it may well be a mystery physically. It's very difficult to crystallize proteins to determine their 3D structure. I think good labs can do maybe 3 or 4 a year. More vexing is the fact that probably a lot of useful targets are on cell membranes, and membrane-bound proteins are usually impossible to crystallize at all.
(2) The solvent has a profound influence on the interaction between macromolecules and possible ligands, so computer simulation of these systems has to take good account of the solvent. But the solvent (water) is small and moves on a femtosecond time-scale, while the interactions of interest are maybe 5 to 10 orders of magnitude slower. That means your computer spends essentially all of its time simulating (useless) solvent behaviour, and very little simulating interesting protein/ligand behaviour. The problem grows exponentially with the size of the system of interest, so it can't be solved even by Moore's Law. Better theoretical models of solvent are needed.
(3) The interactions that govern the dynamics of these systems are strongly many-body, but the calculation of true many-body dynamics is prohibitively expensive. Inevitably simplifying approximations are made, but that tends to reduce the utility of the methods, because the approximations must be validated in a system not too dissimilar from your target system. That means you have to have some independent means of knowing the behaviour of a system not too far from your target system to validate your computer model. That makes them less generalizable and useful than one would hope.
That said, it's an area of vigorous research and much progress. There have been some noteable successes. But the idea in the 80s that by now we'd be routinely designing drugs by computer simulation is sort of like the idea in the 60s that by 2000 we'd have moon colonies and be routinely sending manned spacecraft to Mars. The future is arriving more slowly than we'd hoped. As it always does, except maybe around April 15.