bfio 1.3.5

Simple reading and writing classes for OME tiled tiffs using Bioformats.

BioFormats Input/Output utility (bfio)

This tool is a simplified but powerful interface to the Bioformats java library. It makes use of Cell Profilers python-bioformats package to access the Bioformats library. One of the issues with using the python-bioformats package is reading and writing large image planes (>2GB). The challenge lies in the way Bioformats reads and writes large image planes, using an int value to index the file. To do get around this, files can be read or written in chunks and the classes provided in bfio handle this automatically. The BioWriter class in this package only writes files in the .ome.tif format, and automatically sets the tile sizes to 1024.

Docker containers with all necessary components are available (see Docker Containers section).

Examples

Read an Image

import javabridge
from bfio import BioReader,BioWriter,JARS
from pathlib import Path

# Import javabridge and start the vm
javabridge.start_vm(class_path=JARS)

# Path to bioformats supported image
image_path = Path('path/to/file.ome.tif')

# Create the BioReader object
bf = BioReader(str(image_path))

# Load the full image
image = bf.read_image()

# Only load the first 256x256 pixels, will still load all Z,C,T dimensions
# Note: Images are always 5-dimensional
image = bf.read_image(X=(0,256),Y=(0,256))

# Only load the second channel
image = bf.read_image(C=[1])

# Done executing program, so kill the vm. If the program needs to be run
# again, a new interpreter will need to be spawned to start the vm.
javabridge.kill_vm()

Write an Image

import javabridge
from bfio import BioReader,BioWriter,JARS
from pathlib import Path

# Import javabridge and start the vm
javabridge.start_vm(class_path=JARS)

# Path to bioformats supported image
image_path = Path('path/to/file.ome.tif')

# Create the BioReader object
br = BioReader(str(image_path))

# Load the full image
image = br.read_image()

# Create an image write object, rename the channels
bw = BioWriter(str(image_path.with_name("New_" + image_path.name)),image=image)
bw.channel_names(["Empty","ZO1","Empty"])
bw.write_image(image)
bw.close_image()

# Only save one channel
bw = BioWriter(str(image_path.with_name("New_" + image_path.name)),image=image)
bw.num_c(1)
bw.write_image(image[:,:,0,1,0].reshape((image.shape[0],image.shape[1],1,1,1)))
bw.close_image()

# List the channel names
print(bw.channel_names())

# Done executing program, so kill the vm. If the program needs to be run
# again, a new interpreter will need to be spawned to start the vm.
jutil.kill_vm()

Universal Container Components

All containers contain the follow components:

1. Python 3.6
2. openjdk-8
3. numpy (version 1.18.1)
4. javabridge (version 1.0.18)
5. python-bioformats (version 1.5.2)
6. bfio (version 1.3.5)
7. loci-tools.jar (Version 6.1.0)

Docker Containers

labshare/polus-bfio-util:1.3.5 & labshare/polus-bfio-util:1.3.5-alpine

This container is built on Alpine Linux. This is the smallest bfio container, but also the most difficult to install additional requirements on.

labshare/polus-bfio-util:1.3.5-slim-buster

This container is built on a stripped down version of Debian Buster. This container is larger than the alpine version, but easier to install new Python packages on.

labshare/polus-bfio-util:1.3.5-tensorflow

This container is built on Debian Buster and includes Tensorflow 2.1.0 and all necessary GPU drivers to run Tensorflow on an NVIDIA graphics card.