medstudent2005

Welcome to the Jungle
7+ Year Member
15+ Year Member
Aug 8, 2003
579
0
welle
Visit site
Status
hey guys

i fellow med student told me that his dad, who is an anesthesiologist, mentioned something about artificial intelligence type of radiographic readings, as in, the computer pretty much reads the film, and the radiologist may just have to sign it off or something like that. is this just mad science or is there any truth to this ?


thanks
NT
 

Drakensoul

Senior Member
7+ Year Member
15+ Year Member
Mar 1, 2004
151
0
35
Gainesville, Florida
Visit site
Status
Originally posted by medstudent2005
hey guys

i fellow med student told me that his dad, who is an anesthesiologist, mentioned something about artificial intelligence type of radiographic readings, as in, the computer pretty much reads the film, and the radiologist may just have to sign it off or something like that. is this just mad science or is there any truth to this ?


thanks
NT
We don't even have speech-recognition programs that work with all that great of an accuracy percentage. I can't imagine what it would take for a computer to be able to compare a radio image to some pre-selected 'baseline' image, scan for deviations, convert the deviations into a diagnosis, and be in any way accurate. Seems like far too much room for error. :D
 

Docxter

Senior Member
7+ Year Member
15+ Year Member
Jan 27, 2004
462
3
Visit site
Status
Originally posted by medstudent2005
hey guys

i fellow med student told me that his dad, who is an anesthesiologist, mentioned something about artificial intelligence type of radiographic readings, as in, the computer pretty much reads the film, and the radiologist may just have to sign it off or something like that. is this just mad science or is there any truth to this ?


thanks
NT
There is no computer program that will "read" a CT, xray, or MRI and there won't be one at least in the very near future. However, there are multiple computer programs that are commercially available which "help"radiologists. The vast majority have some segmentation algorithm function which enables them to identify certain target structures as part of a CAD (computer aided detection) scheme. For example, the most widely distributed and advertised software is one that helps with mammography. What it does is that it tries to identify clusters of bright spots (i.e., regions possibly suggestive of microcalcification clusters) and dense regions and dense regions with radiating lines (i.e., regions possibly suggestive of masses or architectural distortions) and highlights them. Then, the radiologist has to look at those highlighted regions and decide whether it is a real finding or a significant finding or not. So in other words, it doesn't read a film, but helps identify areas that "MAY" have something. Again, this software has been extensively tested and is not fail-proof. It, too, has sensitivity threshold and marker thresholds that are adjustable, which increase sensitivity and decrease specificity, or vice versa, just like ROC curvepoints.

There are products that are being developed in the past 15 years and the newest guy in the commercial market is one which helps identify structures that may be pulmonary nodules on a chest x-ray. At every RSNA and SCAR meeting there are hundreds of these segmentation programs being shown as posters or exhibits. These are all useful products that will improve the workflow of radiology.
 
About the Ads

medstudent2005

Welcome to the Jungle
7+ Year Member
15+ Year Member
Aug 8, 2003
579
0
welle
Visit site
Status
doxter man, you are one knowledgeable dude! thanks for that answer...
 

Santiago

Catheter Jockey
7+ Year Member
15+ Year Member
Apr 23, 2003
252
0
Neverland
Visit site
Status
Originally posted by medstudent2005
hey guys

i fellow med student told me that his dad, who is an anesthesiologist, mentioned something about artificial intelligence type of radiographic readings, as in, the computer pretty much reads the film, and the radiologist may just have to sign it off or something like that. is this just mad science or is there any truth to this ?


thanks
NT
CAD & ANN is already being used in mammography and pulmonary nodule evaluation(in nascent stages as of now).And I beleive it has a role in a lot of things.
Check this out
http://radiology.rsnajnls.org/cgi/c...jkey=bde65f42daaf72d9eacbe5b1952cf53f0bbe8ed6
 

medstudent2005

Welcome to the Jungle
7+ Year Member
15+ Year Member
Aug 8, 2003
579
0
welle
Visit site
Status
possibly that's what my friend's dad was referring to...the article was a bit over my head though, but came from the med school i attend so i may have the chance to get more first hand info.

NT
 

Santiago

Catheter Jockey
7+ Year Member
15+ Year Member
Apr 23, 2003
252
0
Neverland
Visit site
Status
hi,
Doxter has explained the whole thing beautifully though.Cant get more lucid than what he said.
 
About the Ads