LeeKamentsky
commented
9 years ago I think the problem is that bioformats is using the wrong reader. It should be using loci.formats.in.OMETiffReader and instead is using loci.formats.in.TiffDelegateReader. This is because bioformats, by default, prevents the ImageReader class from examining the file contents if it can find a reader by looking at the file's extension. The OMETiffReader insists on reading the file before accepting that it is the file's reader, but the TiffDelegateReader does not, so the TiffDelegateReader is chosen as the reader.
We do this to limit the time it takes to choose a reader and to limit the impact of (potentially) 100+ readers each reading the file from disk. This can have a heavy impact for our parent project, CellProfiler, if run on a cluster (We've brought down an Isolon file server before).
You can use the loci.formats.ImageReader class to pick the image reader and override this behavior:
import javabridge
import bioformats.formatreader
javabridge.start_vm(class_path=bioformats.JARS)
from pylab import *
rcParams['figure.figsize'] = (63, 27)
path = r"c:\Temp\output\bfissue23\multi-channel-time-series.ome.tif"
image_reader = bioformats.formatreader.make_image_reader_class()()
# This sets up the readers to read the contents of the file if necessary
image_reader.allowOpenToCheckType(True)
image_reader.setId(path)
n_c = image_reader.getSizeC()
n_t = image_reader.getSizeT()
# Make a fake ImageReader and install the one above inside it
wrapper = bioformats.formatreader.ImageReader(path=path, perform_init=False)
wrapper.rdr = image_reader
print "This file has %d channels and %d timepoints" % (n_c, n_t)
for c in range(n_c):
for t in range(n_t):
subplot(n_c, n_t, c*n_t + t + 1).imshow(wrapper.read(c=c, t=t), cmap="gray")
savefig(r"c:\Temp\output\bfissue23\figure.png")
We've improved the Javabridge considerably since writing python-bioformats. You might consider using the Java directly to access the metadata:
rdr = javabridge.JClassWrapper('loci.formats.in.OMETiffReader')()
rdr.setOriginalMetadataPopulated(True)
clsOMEXMLService = javabridge.JClassWrapper('loci.formats.services.OMEXMLService')
serviceFactory = javabridge.JClassWrapper('loci.common.services.ServiceFactory')()
service = serviceFactory.getInstance(clsOMEXMLService.klass)
metadata = service.createOMEXMLMetadata()
rdr.setMetadataStore(metadata)
rdr.setId(path)
root = metadata.getRoot()
first_image = root.getImage(0)
pixels = first_image.getPixels()
# The plane data isn't in the planes, it's in the tiff data
for idx in range(pixels.sizeOfTiffDataList()):
tiffData = pixels.getTiffData(idx)
c = tiffData.getFirstC().getValue().intValue()
t = tiffData.getFirstT().getValue().intValue()
print "TiffData: c=%d, t=%d" % (c, t)
christofferaberg
Heerpa
First I should commend the creation of this package. I had just about given up (at least temporarily) on using bioformats due to the lack of python support, when I realised this package had been released. Thanks for that!
I do, however, have some issues with actually using the package. I have tried files from two instruments, and in all cases I am unable to read metadata from individual planes. Note that this metadata is readable by the BioFormats library itself, because I can see it using the BioFormats Import plugin in ImageJ.
To take an easy example, consider one of the bioformats sample files, say:
http://www.openmicroscopy.org/Schemas/Samples/2015-01/bioformats-artificial/multi-channel-time-series.ome.tif.zip
This file contains the following metadata
Based on the documentation, I expected that using the code below, I would be able to traverse this tree of information:
As you can see, the code does not give the correct number of channels, nor the correct number of planes or SizeT.
Now, this example is perhaps somewhat artificial (the XML string that is extracted is correct, so I could in principle traverse its tree manually). However, for all other cases I have tried, I have been unable to read any of the information of the individual planes, and
get_plane_count()always returns 0. Am I, somehow, using the package incorrectly?