0

I'm trying to parse the results page from a dicom study. I've tried with pydicom but it only parses the "header" info (patient info, device info, etc.). I'd like to extract the numeric results from the study in text format. In the following example:

Results sample

I'd like to get a text with these lines:

R EIA 1.13m/s 53 0.14m/s 53

R CFA 0.69m/s 53 0.11m/s 53

etc...

Is there any way to parse this data in Python or any other language to text, with some degree of accuracy? I've tried grabbing PixelData with pydicom, dump it to TIFF format and then performing OCR. The results are quite good, but all the 6's and the 3's are confused by 8's.

Thanks!

M

Suever
  • 64,497
  • 14
  • 82
  • 101
Neurus
  • 657
  • 4
  • 27
  • have you checked https://code.google.com/p/pydicom/wiki/WorkingWithPixelData using NumPy ? – sk11 Sep 23 '14 at 06:56
  • thanks sk11, I did try but didn't get too far... I looked online for a specific example with numpy and PixelData, but couldn't find anything. – Neurus Sep 24 '14 at 05:05
  • is this values saved as image in dicom file ? beacuze this values is not dicom standered , u can acieve this if it saved as PDF inside Dicom File ,or as an SR report – Ahmed Hussein Oct 02 '14 at 01:27
  • aslo can u share sample anonymus study so i can help – Ahmed Hussein Oct 02 '14 at 01:29
  • @marcelo i am sorry data stored as pixel-data not attached report , you dont have choice but choose an OCR for values Or ,you can ask your device manufacturer about if the modality can produce these values as SR report with the study – Ahmed Hussein Oct 02 '14 at 19:27
  • @ah.adel thanks, yeah, I will end up with OCR... so what I'm going to do is export the pixeldata to PNG (with dcmtk, there is a command called dcmj2pnm) and from there get a good OCR. I'm trying with tesseract, but not getting accurate results. Thanks again, I'll report my findings here. – Neurus Oct 03 '14 at 20:41
  • 1
    @ah.adel hey can you put your same response as an answer? I'll mark it as solved and add a comment with my OCR findings. Thanks again – Neurus Oct 12 '14 at 20:10

1 Answers1

1

as your data stored as pixel-data not attached report , you dont have choice but choose an OCR for values OR you can ask your device manufacturer about if the modality can produce these values as SR/PDF report with the study , some devices do this .

Ahmed Hussein
  • 409
  • 11
  • 24