nd2
.nd2 (Nikon NIS Elements) file reader.
This reader provides a pure python implementation of the Nikon ND2 SDK.
It used to wrap the official SDK with Cython, but has since been completely rewritten to be pure python (for performance, ease of distribution, and maintenance) while retaining complete API parity with the official SDK.
Note: This library is not affiliated with Nikon in any way, but we are grateful for assistance from the SDK developers at Laboratory Imaging.
Features good metadata retrieval, direct to_dask and to_xarray options
for lazy and/or annotated arrays, and output to OME-TIFF.
This library is tested against many nd2 files with the goal of maximizing compatibility and data extraction. (If you find an nd2 file that fails in some way, please open an issue with the file!)
:book: Documentation
install
pip install nd2
or from conda:
conda install -c conda-forge nd2
Legacy nd2 file support
Legacy nd2 (JPEG2000) files are also supported, but require imagecodecs. To
install with support for these files use the legacy extra:
pip install nd2[legacy]
Faster XML parsing
Much of the metadata in the file stored as XML. If found in the environment,
nd2 will use lxml which is much faster
than the built-in xml module. To install with support for lxml use:
pip install nd2 lxml
Usage and API
Full API documentation is available at https://tlambert03.github.io/nd2
Quick summary below:
import nd2
import numpy as np
my_array = nd2.imread('some_file.nd2') # read to numpy array
my_array = nd2.imread('some_file.nd2', dask=True) # read to dask array
my_array = nd2.imread('some_file.nd2', xarray=True) # read to xarray
my_array = nd2.imread('some_file.nd2', xarray=True, dask=True) # read to dask-xarray
# or open a file with nd2.ND2File
f = nd2.ND2File('some_file.nd2')
# (you can also use nd2.ND2File() as a context manager)
with nd2.ND2File('some_file.nd2') as ndfile:
print(ndfile.metadata)
...
# ATTRIBUTES: # example output
f.path # 'some_file.nd2'
f.shape # (10, 2, 256, 256)
f.ndim # 4
f.dtype # np.dtype('uint16')
f.size # 1310720 (total voxel elements)
f.sizes # {'T': 10, 'C': 2, 'Y': 256, 'X': 256}
f.is_rgb # False (whether the file is rgb)
# if the file is RGB, `f.sizes` will have
# an additional {'S': 3} component
# ARRAY OUTPUTS
f.asarray() # in-memory np.ndarray - or use np.asarray(f)
f.to_dask() # delayed dask.array.Array
f.to_xarray() # in-memory xarray.DataArray, with labeled axes/coords
f.to_xarray(delayed=True) # delayed xarray.DataArray
# OME-TIFF OUTPUT (new in v0.10.0)
f.write_tiff('output.ome.tif') # write to ome-tiff file
# see below for examples of these structures
# METADATA # returns instance of ...
f.attributes # nd2.structures.Attributes
f.metadata # nd2.structures.Metadata
f.frame_metadata(0) # nd2.structures.FrameMetadata (frame-specific meta)
f.experiment # List[nd2.structures.ExpLoop]
f.text_info # dict of misc info
f.voxel_size() # VoxelSize(x=0.65, y=0.65, z=1.0)
f.rois # Dict[int, nd2.structures.ROI]
f.binary_data # any binary masks stored in the file. See below.
f.events() # returns tabular "Recorded Data" view from in NIS Elements/Viewer
# with info for each frame in the experiment.
# output is passabled to pandas.DataFrame
f.ome_metadata() # returns metadata as an ome_types.OME object
# (requires ome-types package)
# allll the metadata we can find...
# no attempt made to standardize or parse it
# look in here if you're searching for metadata that isn't exposed in the above
# but try not to rely on it, as it's not guaranteed to be stable
f.unstructured_metadata()
f.close() # don't forget to close when not using a context manager!
f.closed # boolean, whether the file is closed
Metadata structures
These follow the structure of the nikon SDK outputs (where relevant). Here are some example outputs
attributes
Attributes(
bitsPerComponentInMemory=16,
bitsPerComponentSignificant=16,
componentCount=2,
heightPx=32,
pixelDataType='unsigned',
sequenceCount=60,
widthBytes=128,
widthPx=32,
compressionLevel=None,
compressionType=None,
tileHeightPx=None,
tileWidthPx=None,
channelCount=2
)
metadata
Note: the metadata for legacy (JPEG2000) files will be a plain unstructured dict.
Metadata(
contents=Contents(channelCount=2, frameCount=60),
channels=[
Channel(
channel=ChannelMeta(
name='Widefield Green',
index=0,
color=Color(r=91, g=255, b=0, a=1.0),
emissionLambdaNm=535.0,
excitationLambdaNm=None
),
loops=LoopIndices(NETimeLoop=None, TimeLoop=0, XYPosLoop=1, ZStackLoop=2),
microscope=Microscope(
objectiveMagnification=10.0,
objectiveName='Plan Fluor 10x Ph1 DLL',
objectiveNumericalAperture=0.3,
zoomMagnification=1.0,
immersionRefractiveIndex=1.0,
projectiveMagnification=None,
pinholeDiameterUm=None,
modalityFlags=['fluorescence']
),
volume=Volume(
axesCalibrated=[True, True, True],
axesCalibration=[0.652452890023035, 0.652452890023035, 1.0],
axesInterpretation=(
<AxisInterpretation.distance: 'distance'>,
<AxisInterpretation.distance: 'distance'>,
<AxisInterpretation.distance: 'distance'>
),
bitsPerComponentInMemory=16,
bitsPerComponentSignificant=16,
cameraTransformationMatrix=[-0.9998932296054086, -0.014612644841559427, 0.014612644841559427, -0.9998932296054086],
componentCount=1,
componentDataType='unsigned',
voxelCount=[32, 32, 5],
componentMaxima=[0.0],
componentMinima=[0.0],
pixelToStageTransformationMatrix=None
)
),
Channel(
channel=ChannelMeta(
name='Widefield Red',
index=1,
color=Color(r=255, g=85, b=0, a=1.0),
emissionLambdaNm=620.0,
excitationLambdaNm=None
),
loops=LoopIndices(NETimeLoop=None, TimeLoop=0, XYPosLoop=1, ZStackLoop=2),
microscope=Microscope(
objectiveMagnification=10.0,
objectiveName='Plan Fluor 10x Ph1 DLL',
objectiveNumericalAperture=0.3,
zoomMagnification=1.0,
immersionRefractiveIndex=1.0,
projectiveMagnification=None,
pinholeDiameterUm=None,
modalityFlags=['fluorescence']
),
volume=Volume(
axesCalibrated=[True, True, True],
axesCalibration=[0.652452890023035, 0.652452890023035, 1.0],
axesInterpretation=(
<AxisInterpretation.distance: 'distance'>,
<AxisInterpretation.distance: 'distance'>,
<AxisInterpretation.distance: 'distance'>
),
bitsPerComponentInMemory=16,
bitsPerComponentSignificant=16,
cameraTransformationMatrix=[-0.9998932296054086, -0.014612644841559427, 0.014612644841559427, -0.9998932296054086],
componentCount=1,
componentDataType='unsigned',
voxelCount=[32, 32, 5],
componentMaxima=[0.0],
componentMinima=[0.0],
pixelToStageTransformationMatrix=None
)
)
]
)
experiment
[
TimeLoop(
count=3,
nestingLevel=0,
parameters=TimeLoopParams(
startMs=0.0,
periodMs=1.0,
durationMs=0.0,
periodDiff=PeriodDiff(avg=16278.339965820312, max=16411.849853515625, min=16144.830078125)
),
type='TimeLoop'
),
XYPosLoop(
count=4,
nestingLevel=1,
parameters=XYPosLoopParams(
isSettingZ=True,
points=[
Position(stagePositionUm=[26950.2, -1801.6000000000001, 498.46000000000004], pfsOffset=None, name=None),
Position(stagePositionUm=[31452.2, -1801.6000000000001, 670.7], pfsOffset=None, name=None),
Position(stagePositionUm=[35234.3, 2116.4, 664.08], pfsOffset=None, name=None),
Position(stagePositionUm=[40642.9, -3585.1000000000004, 555.12], pfsOffset=None, name=None)
]
),
type='XYPosLoop'
),
ZStackLoop(count=5, nestingLevel=2, parameters=ZStackLoopParams(homeIndex=2, stepUm=1.0, bottomToTop=True, deviceName='Ti2 ZDrive'), type='ZStackLoop')
]
rois
ROIs found in the metadata are available at ND2File.rois, which is a
dict of nd2.structures.ROI objects, keyed by the ROI ID:
{
1: ROI(
id=1,
info=RoiInfo(
shapeType=<RoiShapeType.Rectangle: 3>,
interpType=<InterpType.StimulationROI: 4>,
cookie=1,
color=255,
label='',
stimulationGroup=0,
scope=1,
appData=0,
multiFrame=False,
locked=False,
compCount=2,
bpc=16,
autodetected=False,
gradientStimulation=False,
gradientStimulationBitDepth=0,
gradientStimulationLo=0.0,
gradientStimulationHi=0.0
),
guid='{87190352-9B32-46E4-8297-C46621C1E1EF}',
animParams=[
AnimParam(
timeMs=0.0,
enabled=1,
centerX=-0.4228425369685782,
centerY=-0.5194951478743071,
centerZ=0.0,
rotationZ=0.0,
boxShape=BoxShape(
sizeX=0.21256931608133062,
sizeY=0.21441774491682075,
sizeZ=0.0
),
extrudedShape=ExtrudedShape(sizeZ=0, basePoints=[])
)
]
),
...
}
text_info
{
'capturing': 'Flash4.0, SN:101412\r\nSample 1:\r\n Exposure: 100 ms\r\n Binning: 1x1\r\n Scan Mode: Fast\r\nSample 2:\r\n Exposure: 100 ms\r\n Binning: 1x1\r\n Scan Mode: Fast',
'date': '9/28/2021 9:41:27 AM',
'description': 'Metadata:\r\nDimensions: T(3) x XY(4) x λ(2) x Z(5)\r\nCamera Name: Flash4.0, SN:101412\r\nNumerical Aperture: 0.3\r\nRefractive Index: 1\r\nNumber of Picture Planes: 2\r\nPlane #1:\r\n Name: Widefield Green\r\n Component Count: 1\r\n Modality: Widefield Fluorescence\r\n Camera Settings: Exposure: 100 ms\r\n Binning: 1x1\r\n Scan Mode: Fast\r\n Microscope Settings: Nikon Ti2, FilterChanger(Turret-Lo): 3 (FITC)\r\n Nikon Ti2, Shutter(FL-Lo): Open\r\n Nikon Ti2, Shutter(DIA LED): Closed\r\n Nikon Ti2, Illuminator(DIA): Off\r\n Nikon Ti2, Illuminator(DIA) Iris intensity: 3.0\r\n Analyzer Slider: Extracted\r\n Analyzer Cube: Extracted\r\n Condenser: 1 (Shutter)\r\n PFS, state: On\r\n PFS, offset: 7959\r\n PFS, mirror: Inserted\r\n PFS, Dish Type: Glass\r\n Zoom: 1.00x\r\n Sola, Shutter(Sola): Active\r\n Sola, Illuminator(Sola) Voltage: 100.0\r\nPlane #2:\r\n Name: Widefield Red\r\n Component Count: 1\r\n Modality: Widefield Fluorescence\r\n Camera Settings: Exposure: 100 ms\r\n Binning: 1x1\r\n Scan Mode: Fast\r\n Microscope Settings: Nikon Ti2, FilterChanger(Turret-Lo): 4 (TRITC)\r\n Nikon Ti2, Shutter(FL-Lo): Open\r\n Nikon Ti2, Shutter(DIA LED): Closed\r\n Nikon Ti2, Illuminator(DIA): Off\r\n Nikon Ti2, Illuminator(DIA) Iris intensity: 1.5\r\n Analyzer Slider: Extracted\r\n Analyzer Cube: Extracted\r\n Condenser: 1 (Shutter)\r\n PFS, state: On\r\n PFS, offset: 7959\r\n PFS, mirror: Inserted\r\n PFS, Dish Type: Glass\r\n Zoom: 1.00x\r\n Sola, Shutter(Sola): Active\r\n Sola, Illuminator(Sola) Voltage: 100.0\r\nTime Loop: 3\r\n- Equidistant (Period 1 ms)\r\nZ Stack Loop: 5\r\n- Step: 1 µm\r\n- Device: Ti2 ZDrive',
'optics': 'Plan Fluor 10x Ph1 DLL'
}
binary_data
This property returns an nd2.BinaryLayers object representing all of the
binary masks in the nd2 file.
A nd2.BinaryLayers object is a sequence of individual nd2.BinaryLayer
objects (one for each binary layer found in the file). Each BinaryLayer in
the sequence is a named tuple that has, among other things, a name attribute,
and a data attribute that is list of numpy arrays (one for each frame in the
experiment) or None if the binary layer had no data in that frame.
The most common use case will be to cast either the entire BinaryLayers object
or an individual BinaryLayer to a numpy.ndarray:
>>> import nd2
>>> nd2file = nd2.ND2File('path/to/file.nd2')
>>> binary_layers = nd2file.binary_data
# The output array will have shape
# (n_binary_layers, *coord_shape, *frame_shape).
>>> np.asarray(binary_layers)
For example, if the data in the nd2 file has shape (nT, nZ, nC, nY, nX), and
there are 4 binary layers, then the output of np.asarray(nd2file.binary_data) will
have shape (4, nT, nZ, nY, nX). (Note that the nC dimension is not present
in the output array, and the binary layers are always in the first axis).
You can also cast an individual BinaryLayer to a numpy array:
>>> binary_layer = binary_layers[0]
>>> np.asarray(binary_layer)
events()
This property returns the tabular data reported in the Image Properties > Recorded Data tab of the NIS Viewer.
(There will be a column for each tag in the CustomDataV2_0 section of
custom_data above, as well as any additional events found in the metadata)
The format of the return type data is controlled by the orient argument:
'records': list of dicts -[{column -> value}, ...](default)'dict': dict of dicts -{column -> {index -> value}, ...}'list': dict of lists -{column -> [value, ...]}
Not every column header appears in every event, so when orient is either
'dict' or 'list', float('nan') will be inserted to maintain a consistent
length for each column.
# with `orient='records'` (DEFAULT)
[
{
'Time [s]': 1.32686654,
'Z-Series': -2.0,
'Exposure Time [ms]': 100.0,
'PFS Offset': 0,
'PFS Status': 0,
'X Coord [µm]': 31452.2,
'Y Coord [µm]': -1801.6,
'Z Coord [µm]': 552.74,
'Ti2 ZDrive [µm]': 552.74
},
{
'Time [s]': 1.69089657,
'Z-Series': -1.0,
'Exposure Time [ms]': 100.0,
'PFS Offset': 0,
'PFS Status': 0,
'X Coord [µm]': 31452.2,
'Y Coord [µm]': -1801.6,
'Z Coord [µm]': 553.74,
'Ti2 ZDrive [µm]': 553.74
},
{
'Time [s]': 2.04194662,
'Z-Series': 0.0,
'Exposure Time [ms]': 100.0,
'PFS Offset': 0,
'PFS Status': 0,
'X Coord [µm]': 31452.2,
'Y Coord [µm]': -1801.6,
'Z Coord [µm]': 554.74,
'Ti2 ZDrive [µm]': 554.74
},
{
'Time [s]': 2.38194662,
'Z-Series': 1.0,
'Exposure Time [ms]': 100.0,
'PFS Offset': 0,
'PFS Status': 0,
'X Coord [µm]': 31452.2,
'Y Coord [µm]': -1801.6,
'Z Coord [µm]': 555.74,
'Ti2 ZDrive [µm]': 555.74
},
{
'Time [s]': 2.63795663,
'Z-Series': 2.0,
'Exposure Time [ms]': 100.0,
'PFS Offset': 0,
'PFS Status': 0,
'X Coord [µm]': 31452.2,
'Y Coord [µm]': -1801.6,
'Z Coord [µm]': 556.74,
'Ti2 ZDrive [µm]': 556.74
}
]
# with `orient='list'`
{
'Time [s]': array([1.32686654, 1.69089657, 2.04194662, 2.38194662, 2.63795663]),
'Z-Series': array([-2., -1., 0., 1., 2.]),
'Exposure Time [ms]': array([100., 100., 100., 100., 100.]),
'PFS Offset': array([0, 0, 0, 0, 0], dtype=int32),
'PFS Status': array([0, 0, 0, 0, 0], dtype=int32),
'X Coord [µm]': array([31452.2, 31452.2, 31452.2, 31452.2, 31452.2]),
'Y Coord [µm]': array([-1801.6, -1801.6, -1801.6, -1801.6, -1801.6]),
'Z Coord [µm]': array([552.74, 553.74, 554.74, 555.74, 556.74]),
'Ti2 ZDrive [µm]': array([552.74, 553.74, 554.74, 555.74, 556.74])
}
# with `orient='dict'`
{
'Time [s]': {0: 1.32686654, 1: 1.69089657, 2: 2.04194662, 3: 2.38194662, 4: 2.63795663},
'Z-Series': {0: -2.0, 1: -1.0, 2: 0.0, 3: 1.0, 4: 2.0},
'Exposure Time [ms]': {0: 100.0, 1: 100.0, 2: 100.0, 3: 100.0, 4: 100.0},
'PFS Offset []': {0: 0, 1: 0, 2: 0, 3: 0, 4: 0},
'PFS Status []': {0: 0, 1: 0, 2: 0, 3: 0, 4: 0},
'X Coord [µm]': {0: 31452.2, 1: 31452.2, 2: 31452.2, 3: 31452.2, 4: 31452.2},
'Y Coord [µm]': {0: -1801.6, 1: -1801.6, 2: -1801.6, 3: -1801.6, 4: -1801.6},
'Z Coord [µm]': {0: 552.74, 1: 553.74, 2: 554.74, 3: 555.74, 4: 556.74},
'Ti2 ZDrive [µm]': {0: 552.74, 1: 553.74, 2: 554.74, 3: 555.74, 4: 556.74}
}
You can pass the output of events() to pandas.DataFrame:
In [1]: pd.DataFrame(nd2file.events())
Out[1]:
Time [s] Z-Series Exposure Time [ms] PFS Offset PFS Status [] X Coord [µm] Y Coord [µm] Z Coord [µm] Ti2 ZDrive [µm]
0 1.326867 -2.0 100.0 0 0 31452.2 -1801.6 552.74 552.74
1 1.690897 -1.0 100.0 0 0 31452.2 -1801.6 553.74 553.74
2 2.041947 0.0 100.0 0 0 31452.2 -1801.6 554.74 554.74
3 2.381947 1.0 100.0 0 0 31452.2 -1801.6 555.74 555.74
4 2.637957 2.0 100.0 0 0 31452.2 -1801.6 556.74 556.74
5 8.702229 -2.0 100.0 0 0 31452.2 -1801.6 552.70 552.70
6 9.036269 -1.0 100.0 0 0 31452.2 -1801.6 553.70 553.70
7 9.330319 0.0 100.0 0 0 31452.2 -1801.6 554.68 554.68
8 9.639349 1.0 100.0 0 0 31452.2 -1801.6 555.70 555.70
9 9.906369 2.0 100.0 0 0 31452.2 -1801.6 556.64 556.64
10 11.481439 -2.0 100.0 0 0 31452.2 -1801.6 552.68 552.68
11 11.796479 -1.0 100.0 0 0 31452.2 -1801.6 553.68 553.68
12 12.089479 0.0 100.0 0 0 31452.2 -1801.6 554.68 554.68
13 12.371539 1.0 100.0 0 0 31452.2 -1801.6 555.68 555.68
14 12.665469 2.0 100.0 0 0 31452.2 -1801.6 556.68 556.68
ome_metadata()
See the ome-types documentation for details on
the OME type returned by this method.
In [1]: ome = nd2file.ome_metadata()
In [2]: print(ome)
OME(
instruments=[<1 Instrument>],
images=[<1 Image>],
creator='nd2 v0.7.1'
)
In [3]: print(ome.to_xml())
<OME xmlns="http://www.openmicroscopy.org/Schemas/OME/2016-06"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://www.openmicroscopy.org/Schemas/OME/2016-06 http://www.openmicroscopy.org/Schemas/OME/2016-06/ome.xsd"
Creator="nd2 v0.7.1.dev2+g4ea166e.d20230709">
<Instrument ID="Instrument:0">
<Detector Model="Hamamatsu Dual C14440-20UP" SerialNumber="Hamamatsu Dual C14440-20UP" ID="Detector:0"/>
</Instrument>
<Image ID="Image:0" Name="test39">
<AcquisitionDate>2023-07-08T09:30:55</AcquisitionDate>
...
Contributing / Development
To test locally and contribute. Clone this repo, then:
pip install -e .[dev]
To download sample data:
pip install requests
python scripts/download_samples.py
then run tests:
pytest
(and feel free to open an issue if that doesn't work!)
alternatives
Here are some other nd2 readers that I know of, though many of them are unmaintained:
- pims_nd2 - pims-based reader. ctypes wrapper around the v9.00 (2015) SDK
- nd2reader - pims-based reader, using reverse-engineered file headers. mostly tested on files from NIS Elements 4.30.02
- nd2file - another pure-python, chunk map reader, unmaintained?
- pyND2SDK - windows-only cython wrapper around the v9.00 (2015) SDK. not on PyPI
The motivating factors for this library were:
- support for as many nd2 files as possible, with a large test suite an and emphasis on correctness
- pims-independent delayed reader based on dask
- axis-associated metadata via xarray
1 maintainer
Release files for nd2 0.11.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| nd2-0.11.3.tar.gz | 880.6 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| nd2-0.11.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 982.7 kB
Release files / nd2-0.11.3.tar.gz
| Download URL | nd2-0.11.3.tar.gz |
|---|---|
| Size | 880.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
c153fccee87ef3df27accd18cf813502ceb19289de266a61cc29a5346c43fc91
|
|
BLAKE2b-256 checksum How to use checksums |
e3e72bf7dcb2c011aad251d7041ff592c8ceab30fa6d5c40b654434f0e25e5ba
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Mar 27, 2026.
Transparency logRelease files / nd2-0.11.3-py3-none-any.whl
| Download URL | nd2-0.11.3-py3-none-any.whl |
|---|---|
| Size | 102.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
5cf735dc02bb7b8fb400f657f8c04e73f51c28c1032c644a7746948c13be38f9
|
|
BLAKE2b-256 checksum How to use checksums |
79514f3af9658395bf5549b424b8620df790ba60513d0c093222a55ec5cbb720
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Mar 27, 2026.
Transparency log
