Release Notes

Released: 2026-03-19

New & Noteworthy

• New data aggregation with heat map view for well plates directly in arivis Pro
• Export aggregation settings to arivis Hub and get the same results as in arivis Pro at a larger scale
• Preview feature: enable "Formula-based feature creation" to create new custom features using a mathematical formula

Multi Run & Aggregation (Batch license required)

• new: Aggregate Features operation in Analysis Pipeline
• new: pre-define statistical features as needed in the Aggregate Features operation (e.g. count, min, max, mean, ratio ...)
• new: use Multi Run in the Analysis Panel to run the pipeline and aggregate object features for multiple wells or fields of a well plate document
• new: Data Aggregation window to explore the aggregated values for the well plate
• new: show distribution of an aggregated value as heat map for the well plate in Data Aggregation window
• new: export pipelines with Aggregate Features settings to arivis Hub
• new: export aggregated results as XLS or CSV
• new: Multi Run adds a new tag "Multi Run" to allow filtering for the resulting objects of the same run easily
• improved: current image type is preselected for Multi Run

Analysis

• new: Formula-based custom feature allows to compute a feature based on a freely definable mathematical formula (Preview Feature)
• new: use values from referenced objects in Formula-based features to create e.g. ratio to parent
• new: 3D Feret Max object feature
• new: Maximum Inscribed Circle/Sphere features for filled objects
• improved: DL Reconstruction and DL segmenter remember model specifics in pipeline to recover custom settings in case the model is missing
• improved: allow duplicate channels to be used for DL Reconstruction or DL segmenter
• improved: several operations which export data (e.g as XLS or CSV) do better validation for the configures export name or path
• improved: don't try to use CUDA if support ist not installed or GPU is missing or disabled
• improved: determinism of Cellpose results on different machines/hardwares (please be aware: the results could slightly differ from 4.4 pipelines)
• improved: performance of Cellpose and DL inference (especially if running on CPU)
• improved: updated Matryoshka.python sample scrip
• change: using ONNX Runtime 1.23 and CUDA 12.9 now
• change: some demo data and the related sample pipelines have been updated
• fix: sometimes Cellpose operation did not deliver the same results as in previous versions of arivis Pro
• fix: some sample pipelines were using deprecated operations
• fix: DL Trainer had some issues with unicode characters in dataset names

4D Viewer

• improved: settings for export segments as surface use the same surface quality (and smoothing) as in the viewer
• improved: storyboards for data with more than 8 channels
• improved: show preview icon in viewer label while isosurface extraction dialog is open

2D Viewer

• improved: selected objects are visualized with a yellow outline now in the 2D Viewer (similar to the 4D Viewer)
• improved: object selection in 2D Viewer by left click + drag a rectangle is now updated live (Preview Feature)
• improved: left click + drag to select objects in the 2D Viewer can be started from everywhere (even inside an object)

Import & Data Handling

• improved: consider the PlateId for naming on import well plate datasets
• fix: wrong tile positions for PE well plate data without orientation matrix specified in the metadata
• fix: import problems for data from PE Acapella
• fix: for some CZI files it was not possible to hide some channels
• fix: set time does not work properly for some CZI files

General

• new: a link to request a trial license is available right from the License Selection dialog
• new: arivis Pro can be installed from the ZEISS Microscopy Installer (ZMI)
• change: arivis Pro stand alone installer provides Docker Desktop version 4.56.0 now

System requirements

• change: supported & tested platform is Windows 11
• change: arivis Pro requires a GPU providing CUDA compute capabilities ≥ 7.5 for ML/DL acceleration 
• change: NVIDIA GPU minimum driver version is 576.02 
• change: arivis Pro stand-alone installer provides Docker Desktop version 4.56.0 
• change: Docker based instance segmentation requires NVIDIA GPU with CUDA Compute Capability ≥ 7.0 now

Misc

• improved: several updates & fixes for the help content
• improved: updated version of the ZEISS Help Viewer
• fix: potential crashes running in Windows 11 when operations on very large data sets run in parallel with many threads
• various smaller bug fixes and improvements

Released: 2025-09-30

New & Noteworthy

• New Multi Run for the Analysis Panel: run the pipeline for multiple wells or fields of a well plate document
• Retain selected tags when switching between wells or image sets
• Accurately detecting cells and nuclei of varying sizes with the new Cellpose-SAM model
• More intuitive rotation in 4D Viewer using the trackball rotation mode

High Content Analysis

• new: run an analysis pipeline for selected well or fields with the new Multi Run mode
• new: interactively select wells and fields for the analysis multi run with a new well picker
• new: show hierarchical state and progress while running a pipeline over multiple fields and wells
• new: explore the results of already finished wells while the pipeline for other wells is still running
• new: stick to the currently selected tag selection while switching between different wells or fields
• new: stick to the current channel visibility while switching between different wells or fields
• new: distinguish different image types for a single well or field (e.g. fluorescence z-stack, bright-field or maximum intensity projection)

Analysis

• new: Convert Traces to Segments operation
• new: Cellpose-SAM (CPSAM) model is available in the Cellpose-based Segmenter operation
• improved: Segment Morphology operation has additional parameters for some methods
• improved: show a warning in the Batch Analysis if output files with the same name would be generated (and would overwrite each other)
• improved: DL reconstruction speed up by better utilization of GPU resources
• improved: better progress while downloading models from arivis Cloud AI Model Store
• improved: error handling for the Export Surfaces operation
• fixed: some minor bugs for the ML Trainer (Objects) and ML Trainer (Image)
• fixed: issues with the objects in the DL Trainer if classes or training has been renamed before drawing
• fixed: potential problem with continue training if multiple image sets are used in the DL Trainer
• fixed: some rounding issues in the max distance field of the Compartments operation
• fixed: potential crash with Magic Wand on CZI data
• fixed: a problem with large planes in CZI data if used with arivis Cloud models in DL Segmenter
• fixed: do not reset the object selection when switching viewer types (while showing objects from the Analysis pipeline)

General

• new: do not reset object filter (e.g. selected tag) when switching between image sets
• new: do not reset channel visibility when switching between image sets (with the same number of channels)
• new: the Preview Features dialog allows you to activate and test early results for newly added features
• new: the arivis Pro help content is provided by the ZEISS help viewer now
• improved: help content layout

Objects

• new: ability to represent feature values as colorful bars normalized to min / max of the Summary (if enabled)
• improved: detection in the AI-assisted mode in the Draw Tool for data zoomed > 100%
• improved: selection of (overlapping) objects in 2D
• improved: help content for object store types and the context menu of the Objects dialog

4D Viewer

• new: Quick Movie Export - export like you see it in the viewer
• new: intuitive rotation of 3D data using the trackball rotation mode
• new: dedicated Rotate tool
• new: improved positioning of clipping planes with handles
• improved: cursor feedback for axis-specific rotation with right mouse and XYZ key
• improved: more anti-aliased selection line
• fixed: prevent artifacts on volume boundaries with inverted opacity curve
• fixed: prevent object render problems with some specific NVIDIA driver versions
• fixed: avoid problems with missing permissions for the clipboard on copy / paste of opacity settings

Data Handling & Import

• new: import image type for well plate data (e.g. fluorescence z-stack, bright-field or maximum intensity projection)
• new: include image type (if available) in the default name for the imported image set
• improved: import tiled LaVision FastTile Blaze data
• improved: normalization of the color range for CZI float datasets
• improved: support added for more compression types of HDF5 blocks (e.g. used by some IMS datasets)
• fixed: potential wrong colors were imported from some Aperio-SVS data
• fixed: wrong pixel size imported from some Aperio-SVS data
• fixed: import of different image types as channels from Yokogawa
• fixed: import of BigDataViewer with non-consecutive setups
• fixed: some CZI appear blank if opened but not imported

Misc

• new: Docker Desktop is now an optional download in the arivis Pro installer
• new: includes a Docker Desktop license covering running arivis Cloud models in arivis Pro
• removed: Extended Focus has been retired
• several additional bug fixes and stability improvements

Released: 2025-04-30

New & Noteworthy

• AI-assisted mode for the Draw Tool - precision and efficiency in manual segmentation workflows.
• Import Document Objects operations allows to copy stored features.

Improvments & Fixes

improved: easier adding CZI data to Batch Analysis

improved: allow download of CZI data from arivis Hub

improved: DL Trainer behavior while drawing at different image sets

fix: object store cache for CZI data uses correct extension now (e.g. filename.czi.cache)

fix: wrong well or field assignment for some scenes in CZI data

fix: scaled analysis could cause artifacts with sparse CZIAnalysis:

fix: DL Trainer - drawn objects were colored wrong (if class was renamed before drawing)

fix: Compartments operations did not store the inputs correctly if the same input was in use multiple times

fix: intensity features are listed under the segments category again

fix: instance segmentation (working on modified channel data) writes results to the correct time point more stable

fix: using trace objects in Python script operation

General

improved: Draw tool visualizes the merging of shapes directly when drawing with the brush* change: arivis License Tool requires .Net 8 now* additional smaller fixes

Released: 2025-03-11

New & Noteworthy

• AI-assisted Spine Tracing: automated spine detection using pre-trained Al models or other methods.
• Enhanced Cell Measurements: expanded measurement capabilities enabling richer feature sets.
• Native CZI Support: open, visualize and analyze CZI data, with no intermediate conversion steps needed.
• Remote Docker: run instance segmentation with the help of a satellite / remote Linux system.

Deep Learning / Machine Learning

• new: DL Trainer preview - verify the trained model right in the DL trainer panel
• new: tile blending option for DL reconstruction operation
• new: option to specify the normalization range to be applied on the data before inference
• new: confidence feature value for instance segmentation results
• new: option to run instance segmentation with the help of a remote Docker engine
• new: arivis Cloud preference page
• new: option to adjust normalization range for Cellpose operation
• improved: improved consistency for CUDA based calculations (e.g. Cellpose)
• improved: better help and guidance for running instance segmentation based on ZEISS arivis Cloud models
• improved: AI model store sorts list by creation time now
• improved: AI models store displays the model's author
• improved: do not show and store arivis Cloud access token in clear text anymore
• improved: order of classes in the DL trainer is reversed (the same as in arivis Cloud now)
• fix: prevent "Could not reach the instance segmentation service" for some specific Docker setups
• fix: ML Segmenter - under some conditions, segments were not found on the last plane
• removed: sync with channel visibility in ML Trainer

Tracing

• new: Spine Tracer operation to detect spines and extend existing traces
• new: AI assisted spine detection in the Spine Tracer operation
• new: spine counting and spine density object features
• new: SWC export respects selection
• fix: Neuron Tracer - make sure all neurites are connected to a cell body (if configured)

Analysis & Objects

• new: SD (standard deviation) added for a lot of features
• new: morphological segment features (convex hull, cross section area, equivalent ellipse projection, maximum inscribed sphere, inscribed disk)
• new: additional 2D segment features (symmetry 2D, Feret projection, convex hull projection)
• improved: Seeded Region Growing supports more and larger seeds
• improved: object styles parent filter allows to select trace type
• improved: updated ITK version (fixes cancellation issues)
• improved: progress in the Store Objects operation during surface pre-calculation
• improved: error handling in the Import Surfaces dialog
• improved: group statistics for traces
• improved: more detailed descriptions for object feature calculations in the help

4D-Viewer

• new: arivis Pro tries to active high performance graphics automatically
• improved: optimized progressive strategy for object visualization
• improved: surface calculation creates smoother objects with less terracing artefacts
• improved: selection of objects behind an active clipping plane
• improved: unified memory limit warnings for traces and segments
• improved: performance for the track tool in 4D
• fix: avoid block artefacts for storyboard movies with time series data

Data Handling & Import

• new: open CZI data natively without the need for import
• new: Thermo Fisher CellInsight CX7 media handler
• improved: background correction performance for large data
• improved: Meta Image Series / ImageXpress import
• improved: support for JPEG2000 compressed TIFF data
• fix: image set names for well plate datasets (e.g. from ImageXpress or Yokogawa)
• fix: potential import problems for 32bit float TIFF data
• fix: problems importing multimodal Yokogawa data
• fix: some VSI were imported with a wrong channel order
• fix: potential problems with export objects as labelled images
• fix: import *.am data from Avizio
• fix: import of stitched LaVision datasets

General

• new: unified Pan tool (previously Drag tool in 2D)
• new: allow panning in 2D and 4D Viewer (independent of the active tool) by middle mouse drag
• fix: prevent selection of non editable objects in the Draw Tool (e.g. track or trace)
• removed: deprecated Python script API functions ImageSet.get_pixel and ImageSet.set_pixel
• switched to .Net 8 * numerous additional bug fixes and stability improvements

Released: 2024-06-18

ZEISS arivis Pro 4.2 (formerly Vision4D) offers users the freedom to select the most suitable image analysis method for their needs. The new release includes enhanced AI-driven tools, 3D analysis capabilities, and large dataset handling. The software allows for unprecedented flexibility to tailor microscopy image analysis workflows to answer research questions.

New & Noteworthy

• Optimized meshing and rendering for unlimited numbers of objects in the 4D Viewer 
• Instance (object-based) segmentation driven by Deep Learning models from arivis Cloud AI Model Store
• Cells and nuclei segmentation in microscopy images using the Cellpose operation 
• High precision preview in the Analysis Pipeline can work on 3D data to produce a more meaningful preview

Other Optimizations

• Analysis can directly show a preview in the 4D Viewer for object creation operations (e.g. Neuron Tracer)
• Machine Learning can auto-optimize the used feature set to improve performance
• Improved handling for huge amounts of objects in the objects table
• Rebranding is completed: The software is now called ZEISS arivis Pro

Released: 2023-11-11

Cellpose Segmenter

• improved: better memory checks to avoid internal errors
• improved: more clear error and warnings
• improved: better description of the available pre-trained models in the help
• change: perform planewise is enabled by default now
• fix: handling of results for time points other than the first one (in combination with voxel filters)
• fix: normalization for data with bias in min intensity value
• fix: other small fixes for the Cellpose Segmenter

Import

• improved: importing channel names for XDCE data
• fix: import problems for some ND2 time series

General

• improved: faster free consumed license after quit of the application