anybioimage

Interactive bioimage viewer widget for Jupyter and marimo notebooks. Built on anywidget, it supports multi-dimensional images, multi-channel composites, mask overlays, annotation tools, and HCS plate navigation.

Installation

uv pip install anybioimage

# With all recommended dependencies (excludes SAM/PyTorch)
uv pip install "anybioimage[all]"

# With SAM support (Python 3.10–3.12, requires PyTorch)
uv pip install "anybioimage[complete]"

Quick Start

Minimal example (NumPy only)

import numpy as np
from anybioimage import BioImageViewer

viewer = BioImageViewer()
viewer.set_image(np.random.randint(0, 255, (512, 512), dtype=np.uint8))
viewer  # displays in notebook

Jupyter

from anybioimage import BioImageViewer
from bioio import BioImage
import bioio_tifffile

viewer = BioImageViewer()
viewer.set_image(BioImage("image.tif", reader=bioio_tifffile.Reader))
viewer  # renders inline

marimo

import marimo as mo
from anybioimage import BioImageViewer
from bioio import BioImage
import bioio_tifffile

viewer = BioImageViewer()
viewer.set_image(BioImage("image.tif", reader=bioio_tifffile.Reader))
mo.ui.anywidget(viewer)

Features

Multi-dimensional images

Supports 5D arrays (TCZYX: Time, Channel, Z-stack, Y, X) with sliders for T, Z, and per-channel controls. Pass a BioImage object for lazy loading — efficient for large TIFF and OME-Zarr files.

from bioio import BioImage
import bioio_tifffile
import bioio_ome_zarr

img = BioImage("image.tif",  reader=bioio_tifffile.Reader)
img = BioImage("image.zarr", reader=bioio_ome_zarr.Reader)
viewer.set_image(img)  # activates T/Z sliders, per-channel LUT controls

Multi-channel composites

Each channel has independent color, brightness/contrast (LUT), and visibility controls via the Layers panel in the toolbar. Channel settings can also be set programmatically:

# Access and modify channel settings
settings = list(viewer._channel_settings)
settings[0] = {**settings[0], "name": "DAPI", "color": "#0000ff"}
viewer._channel_settings = settings

Mask overlays

Add segmentation masks as overlay layers with configurable color, opacity, and contour rendering:

viewer.add_mask(labels, name="Nuclei", color="#ff0000", opacity=0.5)
viewer.add_mask(cells, name="Cells", color="#00ff00", contours_only=True)

# Manage masks
viewer.update_mask_settings(mask_id, opacity=0.3)
viewer.remove_mask(mask_id)
viewer.clear_masks()

HCS plate support

Load OME-Zarr HCS plates with well and FOV navigation dropdowns built into the widget:

viewer = BioImageViewer()
viewer.set_plate("plate.zarr")
viewer  # shows Well / FOV dropdowns

Annotation tools

Tool	Shortcut	Description
Pan	P	Navigate and zoom
Select	V	Select annotations; Delete to remove
Rectangle	R	Draw bounding boxes
Polygon	G	Click vertices, double-click to close
Point	O	Place point markers

Export annotations as DataFrames:

viewer.rois_df      # rectangles: id, x, y, width, height
viewer.polygons_df  # polygons: id, points, num_vertices
viewer.points_df    # points: id, x, y

# Synced widget payload, useful in Jupyter:
viewer.annotations["polygons"]
viewer.value["polygons"]

SAM integration

Automatic segmentation with Segment Anything Model when drawing rectangles or placing points:

viewer.enable_sam(model_type="mobile_sam")  # ~40 MB, fastest
viewer.enable_sam(model_type="sam_b")       # SAM base, ~375 MB

Requires uv pip install "anybioimage[sam]" (Python 3.10–3.12).

Optional dependencies

Extra	Installs	Use case
bioio	bioio, bioio-tifffile	TIFF / OME-Zarr loading
contours	scipy	Contour-only mask rendering
sam	ultralytics (PyTorch)	SAM segmentation
all	bioio + contours	Recommended (no PyTorch)
complete	all + sam	Everything

License

MIT