Forums

medical imaging is the vanguard of signal processing in 21st

Started by George Orwell January 30, 2008
On 1 Feb, 19:04, aruzinsky <aruzin...@general-cathexis.com> wrote:

> Only a small fraction of the people who practice medical imaging know > what they are looking at. &#2013266080;Since nobody previously mentioned that > typical medical images are cartoon quality, I will.
Many years ago I attended a DSP conference where a presentation was given on EKG, heart-beat recordings. I don't remember the exact subject of the presentation, but it *may* have been signal recording, compression and transmission for remote diagnostics; some paramedic scans a patient and a cardiologist on a hospital far away can evaluate the data and suggest treatment. Something like that; the exact context doesn't matter. These students held their presentation, which contained everything except a wrapped instrument: recordings, compressed data, reconstructed data, the context of use for their device -- the lot. After they finished one of the old foxes, a DSP professor who had been involved with the medical industry, stood up an explained quietly that they ought to be very, very careful about comercializing the idea. Lossy data compression methods tend to introduce transient artifacts in the reconstructed data. According to this professor, the cardiologists examine the transients of the heart to come up with a diagnose. So the risk was that some cardiologist would see the compression artifacts without realizing what they were, and start medical examinations (and and maybe even treatments) of the subjects. As far as I know, the ideas for a remote cardiology diagnostics device were not pursued further. Rune
On Feb 1, 10:04 am, aruzinsky <aruzin...@general-cathexis.com> wrote:
> On Jan 31, 3:49 pm, dbd <d...@ieee.org> wrote: > > > On Jan 31, 9:20 am, aruzinsky <aruzin...@general-cathexis.com> wrote: > > ... >
> > > An example of an > > > area without good feedback is the USA judicial system, e.g., recent > > > DNA analysis shows that jurors and judges have a poor history of > > > making correct decisions. The judicial system isn't going to rapidly > > > improve because there is no easy way to view the results of changes. > > > It seems ironic that the feedbackless example is a case where new > > technology identified errors that led the system to adapt by adopting > > new technology. Maybe that's feedforward. > > A judicial system based on statistical decision theory should be > designed by engineers and scientists. That would also screw lawyers > without expertise in math.
Why have you abandoned your claim about feedback and attempted to change the subject to decision theory? And why would anyone want statistical decision theory to have anything to do in the judicial system?
> > > An area with good feedback from experimental results is Natural Image > > > Processing because it is done on a computer and almost everyone knows > > > what they are looking at and what it should look like. However, this > > > would be much better if every image processing journal forced authors > > > to post full sized images on the web because you really can't see much > > > in the tiny pictures printed in journals. > > > If everyone knows 'what they are looking at and what it should look > > like', why should bigger images be a publishing requirement? > > Almost everyone in the image processing community knows when images > are too small but are unlikely to complain for political reasons. In > particular, authors with something to hide want small images so they > can get their useless crap published. For example, in the very > popular paper, Seam Carving for Content-Aware Image Resizing by Shai > Avidan and Ariel Shamir, the images are too small and the authors did > not provide larger images for download in web space. One cannot > properly see the artifacts in these images.
Could you share with us which of the images in the paper you feel are too small and why? http://www.faculty.idc.ac.il/arik/ The authors also provide a video about the technique, without requirement by the publisher of their paper. This seems like another case where your example disagrees with your conclusion. http://www.faculty.idc.ac.il/arik/IMRet-All.mov
> In spite of the > popularity of the paper due to the advertising expertise of the > authors, few people use seam carving. As an unofficial poll, look at > the retouching forumhttp://forums.dpreview.com/forums/forum.asp?forum=1006 > to see how often seam carving is used.
What does the popularity of seam carving have to do with your claim about the size of published images? What does the retouching forum have to do with polls? Why are you changing the subject again?
> >... > Only a small fraction of the people who practice medical imaging know > what they are looking at. Since nobody previously mentioned that > typical medical images are cartoon quality, I will.
Could you cite some actual references to specific images and what you find comic about them? Dale B. Dalrymple
aruzinsky <aruzinsky@general-cathexis.com> wrote in news:e2aa1fdc-f0bf-
4cf9-8989-3aef7ec86f99@q39g2000hsf.googlegroups.com:

> Only a small fraction of the people who practice medical imaging know > what they are looking at.
Your idea of medical imaging, frankly, seems to be an understanding of what goes on at the clinical level. Others, like those that actually read some of the medical imaging literature, have a better feeling for the state of the art and current research directions. -- Scott Reverse name to reply
Rune Allnor <allnor@tele.ntnu.no> wrote in news:db30687f-fde1-412f-a1ab-
af8bb4efa4d7@v29g2000hsf.googlegroups.com:

> As far as I know, the ideas for a remote cardiology > diagnostics device were not pursued further. > >
EMT's are beaming EKGs to ER's every day. -- Scott Reverse name to reply
On Feb 1, 8:39&#2013266080;pm, dbd <d...@ieee.org> wrote:
> On Feb 1, 10:04 am, aruzinsky <aruzin...@general-cathexis.com> wrote: > > > > > > > On Jan 31, 3:49 pm, dbd <d...@ieee.org> wrote: > > > > On Jan 31, 9:20 am, aruzinsky <aruzin...@general-cathexis.com> wrote: > > > ... > > > > > An example of an > > > > area without good feedback is the USA judicial system, e.g., recent > > > > DNA analysis shows that jurors and judges have a poor history of > > > > making correct decisions. &#2013266080;The judicial system isn't going to rapidly > > > > improve because there is no easy way to view the results of changes. > > > > It seems ironic that the feedbackless example is a case where new > > > technology identified errors that led the system to adapt by adopting > > > new technology. Maybe that's feedforward. > > > A judicial system based on statistical decision theory should be > > designed by engineers and scientists. &#2013266080;That would also screw lawyers > > without expertise in math. > > Why have you abandoned your claim about feedback and attempted to > change the subject to decision theory? And why would anyone want > statistical decision theory to have anything to do in the judicial > system? >
I haven't abandoned my claim about feedback. Scientists and engineers have minds trained with empirical feedback and therefore, on average, will perform better in areas with relatively little feedback. For starters, they will look for ways to increase empirical feedback in the judicial arena. Also, human nature is such that people without feedback from reality become pompous and start doing things like wearing robes and white wigs. It is my guess that scientists and engineers would design a judicial system based on decision theory.
> > > > > > > > An area with good feedback from experimental results is Natural Image > > > > Processing because it is done on a computer and almost everyone knows > > > > what they are looking at and what it should look like. &#2013266080;However, this > > > > would be much better if every image processing journal forced authors > > > > to post full sized images on the web because you really can't see much > > > > in the tiny pictures printed in journals. > > > > If everyone knows 'what they are looking at and what it should look > > > like', why should bigger images be a publishing requirement? > > > Almost everyone in the image processing community knows when images > > are too small but are unlikely to complain for political reasons. &#2013266080;In > > particular, authors with something to hide want small images so they > > can get their useless crap published. &#2013266080;For example, in the very > > popular paper, Seam Carving for Content-Aware Image Resizing by Shai > > Avidan and Ariel Shamir, the images are too small and the authors did > > not provide larger images for download in web space. &#2013266080;One cannot > > properly see the artifacts in these images. > > Could you share with us which of the images in the paper you feel are > too small and why?http://www.faculty.idc.ac.il/arik/
All of the pictures in the paper are too small to show most kinks where a seam intersects a straight edge or ridge. Some kinks are visible in Figs. 2 d. and 11 b. but are likely to be overlooked.
> The authors also provide a video about the technique, without > requirement by the publisher of their paper. > This seems like another > case where your example disagrees with your conclusion.http://www.faculty.idc.ac.il/arik/IMRet-All.mov >
I assume this is the same as the flash animation on You Tube. These are also small images. Furthermore, this shows slimy advertising expertise rather than technical writing skill. The paper has poor technical writing. All science and engineering papers should be written in sufficient detail so that others can duplicate and verify their results. The authors do not explicitly state that they did not update the energy function or derivatives between seam removals. This has been a major source of confusion to some people trying duplicate their results. The authors also do not explain how they handled color. And, even if they did, they do not give access to their test images to verify results. I bet the paper was a lot worse before the referees improved it. However, I bet not a single referee tried to implement the algorithm and therefore they have an inferior perspective to someone such as myself who has implemented it.
> > In spite of the > > popularity of the paper due to the advertising expertise of the > > authors, few people use seam carving. &#2013266080;As an unofficial poll, look at > > the retouching forumhttp://forums.dpreview.com/forums/forum.asp?forum=1006 > > to see how often seam carving is used. > > What does the popularity of seam carving have to do with your claim > about the size of published images? What does the retouching forum > have to do with polls? Why are you changing the subject again? >
The paper was popular because the usefulness of the algorithm was overestimated by many readers. If those readers had easy access to all the full sized images, the usefulness of the algorithm would not have been overestimated and the paper would not have been as popular. Thus, the authors unfairly received credit. I think the lead author was hired by Adobe, maybe, for his advertising expertise rather than his technical expertise.
> > >... > > Only a small fraction of the people who practice medical imaging know > > what they are looking at. &#2013266080;Since nobody previously mentioned that > > typical medical images are cartoon quality, I will. > > Could you cite some actual references to specific images and what you > find comic about them? >
Now, eveyone should know that you are not earnest.
On 2 Feb 2008 17:27:31 GMT, Scott Seidman
<namdiesttocs@mindspring.com> wrote:

>Rune Allnor <allnor@tele.ntnu.no> wrote in news:db30687f-fde1-412f-a1ab- >af8bb4efa4d7@v29g2000hsf.googlegroups.com: > >> As far as I know, the ideas for a remote cardiology >> diagnostics device were not pursued further. >> >> > >EMT's are beaming EKGs to ER's every day.
Do you know what sort of radio link equipment they're using to do that? I'm curious. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org
Eric Jacobsen <eric.jacobsen@ieee.org> wrote in 
news:rch9q31ebh1q7hjrpkjn9pc3te9kn4lkav@4ax.com:

> On 2 Feb 2008 17:27:31 GMT, Scott Seidman > <namdiesttocs@mindspring.com> wrote: > >>Rune Allnor <allnor@tele.ntnu.no> wrote in news:db30687f-fde1-412f-
a1ab-
>>af8bb4efa4d7@v29g2000hsf.googlegroups.com: >> >>> As far as I know, the ideas for a remote cardiology >>> diagnostics device were not pursued further. >>> >>> >> >>EMT's are beaming EKGs to ER's every day. > > Do you know what sort of radio link equipment they're using to do > that? I'm curious. > > > Eric Jacobsen > Minister of Algorithms > Abineau Communications > http://www.ericjacobsen.org >
No clue, unfortunately. A google search for wireless ecg is yielding a ton of hits, but its a little tough to separate out transimission of the actual ecg from transmission of emergency-type events. For what its worth, I bet its not a huge compression problem. This stuff has been sent by telephone for decades. You need to have about a 500 Hz sample rate. 30 or 60 seconds of 12-channel data at 500 Hz is not a huge problem. http://cse.spsu.edu/pbobbie/New_pdf/VeryFinalEKGAcquisitionCamReady1.pdf has a bit of a review. -- Scott Reverse name to reply
On Feb 2, 10:34 am, aruzinsky <aruzin...@general-cathexis.com> wrote:
> On Feb 1, 8:39 pm, dbd <d...@ieee.org> wrote: > > > > > On Feb 1, 10:04 am, aruzinsky <aruzin...@general-cathexis.com> wrote: > > > > On Jan 31, 3:49 pm, dbd <d...@ieee.org> wrote: > > > > > On Jan 31, 9:20 am, aruzinsky <aruzin...@general-cathexis.com> wrote: > > > > ... > > > Could you share with us which of the images in the paper you feel are > > too small and why?http://www.faculty.idc.ac.il/arik/ > > All of the pictures in the paper are too small to show most kinks > where a seam intersects a straight edge or ridge. Some kinks are > visible in Figs. 2 d. and 11 b. but are likely to be overlooked. >
Thank you for the examples. I agree with you that these figures are small, but I know of no publishers with editorial standards that allow the space that would be required. Do you?
> > The authors also provide a video about the technique, without > > requirement by the publisher of their paper. > > This seems like another > > case where your example disagrees with your conclusion.http://www.faculty.idc.ac.il/arik/IMRet-All.mov > > I assume this is the same as the flash animation on You Tube. These > are also small images. Furthermore, this shows slimy advertising > expertise rather than technical writing skill.
The authors have provided a larger video version of the Fig 11.b that you cited. You seem to rant about 'slimy advertising' without even looking.
> All science and engineering papers should be > written in sufficient detail so that others can duplicate and verify > their results. >
I'd like to see this too. Do you know any publishers that support this?
> > >... > > > Only a small fraction of the people who practice medical imaging know > > > what they are looking at. Since nobody previously mentioned that > > > typical medical images are cartoon quality, I will. > > > Could you cite some actual references to specific images and what you > > find comic about them? > > Now, eveyone should know that you are not earnest.
On the contrary, This was an earnest attempt to let you demonstrate whether or not you hold yourself the same level of documentation and reference to which you claim others should be held. You are still preaching without practicing. That's the type of advertiser you demonstrate yourself to be. Dale B. Dalrymple
dbd wrote:

   ...

> Thank you for the examples. I agree with you that these figures are > small, but I know of no publishers with editorial standards that allow > the space that would be required. Do you?
The suggestion I remember is that full-size images be made available on the web. ... Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
On 3 Feb, 02:48, Jerry Avins <j...@ieee.org> wrote:
> dbd wrote: > > &#2013266080; &#2013266080;... > > > Thank you for the examples. I agree with you that these figures are > > small, but I know of no publishers with editorial standards that allow > > the space that would be required. Do you? > > The suggestion I remember is that full-size images be made available on > the web.
Why not print zoomed-in detail images at a size which suits the publication? If somebody claims to have a method which preserves a particular type of detail in an image, a journal editor ought to require the article to contain comparision images at a 'suitable' level of detail. Rune