Guest guest Posted April 7, 2008 Report Share Posted April 7, 2008 I'm using a commercial OCR web service. I found that tesseract was not able to cope with text logos and gave pretty poor results in comparison. Synapse also extracts digital results from ocr'd forms and injects them into the patient database. Luckily most of my results do come as HL7 .. it's only the out of towners who come to see me that I have to go to these lengths. Why bother? To help with chronic care management. This is the type of thing I like to view/print out http://synapse-images.s3.amazonaws.com/73C55E9348E30B2419011E4D87DDCDCE.pdf > > I have tried a couple of open source OCR programs on Linux to > see if I could extract the data and automate lab test filing. Our hospital > can generate the lab reports in a fax format and send them. They > refuse to give me an HL7 feed. Since the fax file is actually produced > by the computer, rather than being hand scanned, they are never > skewed and are very consistent in size and type of font and positioning > of the data. I found that the open source OCR programs " GOCR " > and " Tesseract " could do a pretty good job on the computer generated > faxes. I'm pretty sure you could > write a little program that would be able to extract the name, date, and > even the results from those (I haven't done it). But the faxes that were > hand scanned, and not computer generated as a fax without scanning, > gave very poor results with > the open source OCR programs. I imagine some of the commercial > OCR program have better ability to deal with skewed hand scanned > faxes with multiple sizes and type of fonts etc. What OCR " engine " > are you > using in your project? Did you write it yourself or are you using > an open source or commercial OCR program? > > > > > > > > > > > I think that utility would be wonderful, Graham - > > > > I'm fascinated to think it would be possible - I find that even > looking at > > > > all the reports I get, I have trouble scanning for that information. > > > > The names, DOB's, DOS, all scattered differently through the > material. > > > > > > Well, the utility does require that the names, dob and dos are > > > consistently in the same place for a particular provider. If the same > > > provider varies where they put that information, well, that is a much > > > harder task. > > > > -- > > Graham Chiu > > http://www.synapsedirect.com > > Synapse - the use from anywhere EMR. > > > > -- Graham Chiu http://www.synapsedirect.com Synapse - the use from anywhere EMR. Quote Link to comment Share on other sites More sharing options...
Guest guest Posted April 10, 2008 Report Share Posted April 10, 2008 I get a bit of fax spam, not a lot, but enough to be annoying. I'm thinking that it should be easy enough to add some extra functionality to immediately delete faxes with no fax id, or, faxes from known fax spam numbers .. though I believe in the USA you may have laws against fax spam. Also, regarding refill requests, it should be easy enough ( if your fax prints the sender's fax id at the top of the page ), to route those to a particular directory to be dealt with. Lots of possibilities with this technology. > > > I'm using a commercial OCR web service. > > I found that tesseract was not able to cope with text logos and gave > pretty poor results in comparison. > > Synapse also extracts digital results from ocr'd forms and injects > them into the patient database. Luckily most of my results do come as > HL7 .. it's only the out of towners who come to see me that I have to > go to these lengths. > > Why bother? To help with chronic care management. > > This is the type of thing I like to view/print out > > http://synapse-images.s3.amazonaws.com/73C55E9348E30B2419011E4D87DDCDCE.pdf > > > > > > > > I have tried a couple of open source OCR programs on Linux to > > see if I could extract the data and automate lab test filing. Our hospital > > can generate the lab reports in a fax format and send them. They > > refuse to give me an HL7 feed. Since the fax file is actually produced > > by the computer, rather than being hand scanned, they are never > > skewed and are very consistent in size and type of font and positioning > > of the data. I found that the open source OCR programs " GOCR " > > and " Tesseract " could do a pretty good job on the computer generated > > faxes. I'm pretty sure you could > > write a little program that would be able to extract the name, date, and > > even the results from those (I haven't done it). But the faxes that were > > hand scanned, and not computer generated as a fax without scanning, > > gave very poor results with > > the open source OCR programs. I imagine some of the commercial > > OCR program have better ability to deal with skewed hand scanned > > faxes with multiple sizes and type of fonts etc. What OCR " engine " > > are you > > using in your project? Did you write it yourself or are you using > > an open source or commercial OCR program? > > > > > > > > > > > > > > > > I think that utility would be wonderful, Graham - > > > > > I'm fascinated to think it would be possible - I find that even > > looking at > > > > > all the reports I get, I have trouble scanning for that information. > > > > > The names, DOB's, DOS, all scattered differently through the > > material. > > > > > > > > Well, the utility does require that the names, dob and dos are > > > > consistently in the same place for a particular provider. If the same > > > > provider varies where they put that information, well, that is a much > > > > harder task. > > > > > > -- > > > Graham Chiu > > > http://www.synapsedirect.com > > > Synapse - the use from anywhere EMR. > > > > > > > > > > > -- > Graham Chiu > http://www.synapsedirect.com > Synapse - the use from anywhere EMR. > -- Graham Chiu http://www.synapsedirect.com Synapse - the use from anywhere EMR. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You are posting as a guest. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.