arivis Vision4D Release 3.4

Support for Organization-Wide Image Science

With this release, arivis Vision4D will become more useful to more people in your organization. We are happy to share several new features including:
•    An additional licensing selection that makes it easier to distribute tailored solutions to users.
•    Improvements in Machine Learning to help subject matter experts make effective pipelines more quickly.
•    Updated layout of the Analysis Panel helps everyone – novice to pro – to build pipelines.
•    An overhaul of the Vision4D-Python connection that enables programmers to better control, leverage, and extend Vision4D tools in combination with their own.

 

New detailed Functionalities of arivis Vision4D 3.4

License Management

A new License Selection tool enables users to browse and select arivis licenses at launch. This feature blends with our flexible licensing options and tools to provide better management and deployment of licenses at your site. For example, some users may only need a lightweight viewer while others need a full suite of analysis tools. Now they can select the licensing appropriate for their work. License types (local, network, permanent, time expiring, etc.) are indicated graphically and the individual modules covered by a license are listed. Preference can be remembered for future launches by a user.

arivis Vision4D 3.4 License Management

Machine Learning across images

Image variation often confounds the development of a one-size-fits-all analysis strategy. If the training image for Machine Learning lacks the variations present in the other images, the trained operation will fail. To account for troublesome variations, you can now make a bona fide training set, as a single SIS file, and train the pixel classifier across all the images. Each time you paint a new set of pixels, you receive an updated preview. You can move between the images in the training set to check the training and add regions. Because the Machine Learning tool in Vision4D takes hand-drawn regions for training, subject matter experts can directly drive the development of custom filters and segmentation operations that work for their data.

Paint foreground and background objects on a collection of images to make training that works for a batch of images:
arivis Vision4D 3.4 Machine Learning across images

Efficient Machine Learning

The Machine Learning tool now also leverages the SIS multiresolution format to enable faster results. If structures of interest are very big, users can adjust scaling to achieve satisfactory results in a fraction of the time it would take to run at full resolution.

Selection of the scaling is done easily via dropdown and takes effect immediately to speed up training:
arivis Vision4D 3.4 Selection of the scaling is done easily via dropdown
Pixel classifier trained and run on a 5k x 5k x 47 image in just seconds:
arivis Vision4D 3.4 Pixel classifier trained and run on a 5k x 5k x 47 image in just seconds

Easily access powerful analysis and processing

An expansive set of imaging science algorithms is now organized by operation type and subcategory. Each operator has a short description, so users know what it does before they try it. The view of the list is adjustable, so it shows in full or by parts, or is completely hidden. Novice users can quickly browse the high-level categories to find operations appropriate to their goals, while advanced users can always show all operations and use a search tool to find specific ones.

Categories, subcategories, and an example search for operations that work with ‘membranes’:
arivis Vision4D 3.4 Categories, subcategories, and an example search for operations that work with ‘membranes’

Advanced math tools

The latest additions to the Analysis Panel operations include Image Math and Object Math. Image Math is a full set of operations for math with Channels and Time Series. Object Math enables users to Subtract, Intersect, and Merge classes of segmented objects. The new sets of operations can help advanced users create complex measurements including dynamics and overlaps.

Object Math -Subtraction: separation of cell nuclei and cytoplasm objects:
arivis Vision4D 3.4 Object Math -Subtraction (cells – nuclei = cytpoplasm)
 
Object Math - Intersection:  objects resulting from the intersection of two classes of objects:
arivis Vision4D 3.4 Object Math Intersections

Full Python 3 extendibility

Integrated python scripting is now based on Python 3 (with NumPy included) and offers better support for debugging and notably for external Python installations. Now tools in the python world (e.g, SciPy, StarDist, and PyTorch) can be combined with Vision4D to enable programmers to use their existing python scripts while visualizing results and intermediate processing steps in Vision4D. Additionally, programmers can use and apply Deep Learning networks with Vision4D serving as an interface for subject matter experts to do training.

 

infoatarivis.com (CONACT US)