Feature Extraction¶

This sections will cover how to extract regionprops features from an image using ngio, skimage. Moreover we will also write the features to a table in the ome-zarr container.

Step 1: Open the OME-Zarr Container¶

In [1]:

Copied!





import numpy as np
import pandas as pd
from skimage import measure


def extract_features(image: np.ndarray, label: np.ndarray) -> pd.DataFrame:
    """Basic feature extraction using skimage.measure.regionprops_table."""
    label = label.squeeze(-1)  # Remove the channel axis if present
    roi_feat_table = measure.regionprops_table(
        label_image=label,
        intensity_image=image,
        properties=[
            "label",
            "area",
            "mean_intensity",
            "max_intensity",
            "min_intensity",
        ],
    )
    return pd.DataFrame(roi_feat_table)
import numpy as np
import pandas as pd
from skimage import measure


def extract_features(image: np.ndarray, label: np.ndarray) -> pd.DataFrame:
    """Basic feature extraction using skimage.measure.regionprops_table."""
    label = label.squeeze(-1)  # Remove the channel axis if present
    roi_feat_table = measure.regionprops_table(
        label_image=label,
        intensity_image=image,
        properties=[
            "label",
            "area",
            "mean_intensity",
            "max_intensity",
            "min_intensity",
        ],
    )
    return pd.DataFrame(roi_feat_table)

In [2]:

Copied!





from pathlib import Path

from ngio import open_ome_zarr_container
from ngio.utils import download_ome_zarr_dataset

# Download the dataset
download_dir = Path(".").absolute().parent.parent / "data"
hcs_path = download_ome_zarr_dataset("CardiomyocyteTinyMip", download_dir=download_dir)
image_path = hcs_path / "B" / "03" / "0"

# Open the ome-zarr container
ome_zarr = open_ome_zarr_container(image_path)
from pathlib import Path

from ngio import open_ome_zarr_container
from ngio.utils import download_ome_zarr_dataset

# Download the dataset
download_dir = Path(".").absolute().parent.parent / "data"
hcs_path = download_ome_zarr_dataset("CardiomyocyteTinyMip", download_dir=download_dir)
image_path = hcs_path / "B" / "03" / "0"

# Open the ome-zarr container
ome_zarr = open_ome_zarr_container(image_path)

/opt/hostedtoolcache/Python/3.13.11/x64/lib/python3.13/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

Downloading data from 'https://zenodo.org/records/13305156/files/20200812-CardiomyocyteDifferentiation14-Cycle1_mip.zarr.zip' to file '/home/runner/work/ngio/ngio/data/20200812-CardiomyocyteDifferentiation14-Cycle1-tiny-mip.zarr.zip'.

Unzipping contents of '/home/runner/work/ngio/ngio/data/20200812-CardiomyocyteDifferentiation14-Cycle1-tiny-mip.zarr.zip' to '/home/runner/work/ngio/ngio/data/tmp'

Step 2: Setup the inputs¶

In [3]:

Copied!





from ngio.transforms import ZoomTransform

# First we will need the image object and the FOVs table
image = ome_zarr.get_image()

# Get the nuclei label
nuclei = ome_zarr.get_label("nuclei")

# In this example we the image is available at an higher resolution than the nuclei
print(f"Image dimensions: {image.dimensions}, pixel size: {image.pixel_size}")
print(f"Nuclei dimensions: {nuclei.dimensions}, pixel size: {nuclei.pixel_size}")

# We need to setup a transform to resample the nuclei to the image resolution
zoom_transform = ZoomTransform(
    input_image=nuclei,
    target_image=image,
    order="nearest",  # Nearest neighbor interpolation for labels
)
from ngio.transforms import ZoomTransform

# First we will need the image object and the FOVs table
image = ome_zarr.get_image()

# Get the nuclei label
nuclei = ome_zarr.get_label("nuclei")

# In this example we the image is available at an higher resolution than the nuclei
print(f"Image dimensions: {image.dimensions}, pixel size: {image.pixel_size}")
print(f"Nuclei dimensions: {nuclei.dimensions}, pixel size: {nuclei.pixel_size}")

# We need to setup a transform to resample the nuclei to the image resolution
zoom_transform = ZoomTransform(
    input_image=nuclei,
    target_image=image,
    order="nearest",  # Nearest neighbor interpolation for labels
)

Image dimensions: Dimensions(c: 1, z: 1, y: 2160, x: 5120), pixel size: x=0.1625 y=0.1625 z=1.0 t=1.0 space_unit='micrometer' time_unit=None
Nuclei dimensions: Dimensions(z: 1, y: 540, x: 1280), pixel size: x=0.65 y=0.65 z=1.0 t=1.0 space_unit='micrometer' time_unit=None

Step 3: Use the FeatureExtractorIterator to create a feature table¶

In [4]:

Copied!





from ngio.experimental.iterators import FeatureExtractorIterator
from ngio.tables import FeatureTable

iterator = FeatureExtractorIterator(
    input_image=image,
    input_label=nuclei,
    label_transforms=[zoom_transform],
    axes_order=["y", "x", "c"],
)

feat_table = []
for image_data, label_data, roi in iterator.iter_as_numpy():
    print(f"Processing ROI: {roi}")
    roi_feat_table = extract_features(image=image_data, label=label_data)
    feat_table.append(roi_feat_table)

# Concatenate all the dataframes into a single one
feat_table = pd.concat(feat_table)
feat_table = FeatureTable(table_data=feat_table, reference_label="nuclei")
ome_zarr.add_table("nuclei_regionprops", feat_table)
from ngio.experimental.iterators import FeatureExtractorIterator
from ngio.tables import FeatureTable

iterator = FeatureExtractorIterator(
    input_image=image,
    input_label=nuclei,
    label_transforms=[zoom_transform],
    axes_order=["y", "x", "c"],
)

feat_table = []
for image_data, label_data, roi in iterator.iter_as_numpy():
    print(f"Processing ROI: {roi}")
    roi_feat_table = extract_features(image=image_data, label=label_data)
    feat_table.append(roi_feat_table)

# Concatenate all the dataframes into a single one
feat_table = pd.concat(feat_table)
feat_table = FeatureTable(table_data=feat_table, reference_label="nuclei")
ome_zarr.add_table("nuclei_regionprops", feat_table)

Processing ROI: name=None slices=[z: 0.0->1.0, y: 0.0->351.0, x: 0.0->832.0] label=None space='world'

/opt/hostedtoolcache/Python/3.13.11/x64/lib/python3.13/site-packages/anndata/_io/zarr.py:44: UserWarning: Writing zarr v2 data will no longer be the default in the next minor release. v3 data will be written by default. If you are explicitly setting this configuration, consider migrating to the zarr v3 file format.
  f = open_write_group(store)

Sanity Check: Read the Table back¶

In [5]:

Copied!

ome_zarr.get_table("nuclei_regionprops").lazy_frame.collect()
ome_zarr.get_table("nuclei_regionprops").lazy_frame.collect()

Out[5]:

shape: (1_497, 5)

label	area	mean_intensity-0	max_intensity-0	min_intensity-0
i64	f64	f64	f64	f64
1	1360.0	184.579412	268.0	125.0
2	2464.0	273.246753	461.0	132.0
3	1968.0	277.291667	429.0	143.0
4	5120.0	279.035156	413.0	118.0
5	288.0	243.315972	341.0	147.0
…	…	…	…	…
1493	1248.0	269.291667	385.0	108.0
1494	528.0	315.710227	475.0	128.0
1495	608.0	281.544408	377.0	122.0
1496	288.0	261.003472	356.0	122.0
1497	432.0	336.101852	454.0	172.0

In [ ]: