Python's exportOmeTiff and ROI's

[markmendell]

Hi,

I had trouble opening an exported OME-TIFF into ImageJ, specifically this error:

Code: Select all

java.lang.NumberFormatException: For input string: "points[694"
   at sun.misc.FloatingDecimal.readJavaFormatString(FloatingDecimal.java:1222)
   at java.lang.Double.parseDouble(Double.java:510)
   at loci.plugins.util.ROIHandler.parsePoints(ROIHandler.java:927)
   at loci.plugins.util.ROIHandler.openROIs(ROIHandler.java:250)
   at loci.plugins.in.DisplayHandler.displayROIs(DisplayHandler.java:214)
   at loci.plugins.in.Importer.run(Importer.java:92)
   at loci.plugins.LociImporter.run(LociImporter.java:78)
   at ij.IJ.runUserPlugIn(IJ.java:216)
   at ij.IJ.runPlugIn(IJ.java:180)
   at ij.Executer.runCommand(Executer.java:137)
   at ij.Executer.run(Executer.java:66)
   at java.lang.Thread.run(Thread.java:695)

It looks like exportOmeTiff sets the "Points" attributes of "Polygon" elements in the XML to something like:

points[694,88 695,88 695,89 695,90 696,91 697,91 698,90 ...] points1[694,88 695,88 695,89 695,90 696,91 697,91 698,90 ...] points2[694,88 695,88 695,89 695,90 696,91 697,91 698,90 ...] mask[0,0,0...0]

and bioformats is expecting something like

694,88 695,88 695,89 695,90 696,91 697,91 698,90 ...

Here's a relevant part of the upload script I used: view

Here's the uploaded image in OMERO's viewer: view

Here's the broken exported OME-TIFF: download

And here's the working version with fixed Points attributes: download

(Bonus: the python script used to fix the broken image (requires "$ sudo pip install lxml tifffile"; does not preserve tags other than the ome xml):

Code: Select all

import lxml.etree
import tifffile


IMAGE_PATH = "path/to/broken.tiff"


with tifffile.TiffFile(IMAGE_PATH) as tif:
  data = tif.asarray()
  page_0 = tif.pages[0]
  ome_xml_string = page_0.tags['image_description'].value
ome_xml = lxml.etree.fromstring(ome_xml_string)
all_polygons = ome_xml.findall(".//{*}Polygon")
for polygon in all_polygons:
  points = polygon.get("Points")
  i = points.find("[")
  j = points.find("]")
  new_points = points[i+1:j]
  polygon.set("Points", new_points)
new_s = lxml.etree.tostring(ome_xml, encoding="UTF-8", xml_declaration=True)
tifffile.imsave("out.tiff", data, description=new_s)


)

[dgault]

Hi Mark,

Thanks for spotting and reporting this issue as well investigating the problem. This was indeed a bug and as per the OME XML specification the correct points data should be a list of comma separated x,y coordinates seperated by spaces like "x1,y1 x2,y2 x3,y3"

A fix to export the data in the correct format has actually just been implemented recently so should be coming soon to future Omero release.

David