Imaris 10.1 – Trainable AI Image Analysis for Everyone

Optimize your image analysis workflows using AI trainable object detection in Imaris 10.1. Imaris AI pixel classifier is unique, featuring:

Efficient training on thick slices (visualizing biological structures)
Smooth and slick brush training tool
Superfast predictions preview
Automated use on multiple images (with retrain option)
Fast computations for big datasets (100s of GBs)

Boost your Image analysis workflow with AI Segmentation

Organoids

Imaris AI image analysis workflow (pixel classification and object classification) used for quantitative analysis of organoids. Video courtesy of A. Boyreau, CNRS and L. Andrique VoxCell Facility, TBMcore

MRI Images

Imaris AI segmentation helps to segment and visualize greyscale images, such as CT or MRI. Video based on private data.

Cell Biology

Imaris 10.1 AI segmentation workflows for precise segmentation of cellular structures. Video courtesy of J.P. Dompierre & M. Rojo, Université Bordeaux, France.

SEM images

Automated segmentation of some SEM images can be largely improved using AI image segmentation in Imaris. Video courtesy of R. Schekman & A. Gorur, University of California

New Imaris AI Segmentation Assets

Unique training mode

Training your classifier on a thick serial section (which should be adjusted to your structure thickness) largely improves the comfort and efficiency of your brush strokes. Thick slicers can be adjusted to the size of structures, which are otherwise not clearly visible on a single section, such as neurons, microglia, or cell nuclei.

Even better, the slicer can be positioned in any direction perpendicular or oblique to image axes. Imaris Smart Brush picks the right voxels for you, based on what you see in the extended slice. Try it!

Fast Previews

The pixel classifier training previews are generated extremely fast so you can validate your brush strokes in almost no time (depending on the data size of course).

The previews fist appears in the section you are at and in the meantime are gradually built in other sections. It’s possible that you don’t even notice this process as by the time you go through the slices they’ll already be there. As a bonus, we added a fast 3D preview so you can easily check the image peripheries.

Computational Speed Improvements

Imaris still needs to compute Surfaces from the predictions, but this happens very fast and is using all of your CPU cores. Important to say, that for this process you don’t need an expensive graphic card as all computation is done on CPU.

When you add Imaris Big Surface rendering (since Imaris 9) and Insightful Rendering (since Imaris 9.5) – you get an extremely powerful AI segmentation method which enables to create beautiful models from your data and get the quantitative information from them. Faster algorithms and real-time previews = less computing time spent on the task = less energy consumption

AI Segmentation for Multiple Images

One of the things you may want to do is to apply the same classifier to multiple images and to train it on several images. Both are possible. When you store Surfaces creation parameters the classifier is stored with those creation parameters and the training data are stored with the creation parameters.

The stored creation parameters can be used to run batch processing. The stored creation parameters can also be opened on a new image to continue training on that new image.

Imaris 10.1 Extra Features

Objects on one slicer

Have you tried Imaris object visualization on extended sections slicers? The slicer visualization largely helps to set up and evaluate segmentation results in large and thick 3D datasets as well as select or classify interesting objects, which are otherwise occluded by other structures. If you have missed the possibility to visualize multiple objects (Spots, Filaments, Surfaces) on the same slicer to assess their interaction or overlap, you would love this feature in Imaris 10.1.

Imaris 2D

Imaris has a unique capability to render and visualize large 3D datasets (exceeding the computer RAM), enabling smooth rotation in all directions and zooming.

Nevertheless, not all the datasets are 3D – now Imaris perfectly renders your 2D and 2D time datasets on a single plane and offers accurate 2 D measurements.

The new AI Machine Learning Segmentation is very easy to use and very efficient on big 3D data. I was impressed by its speed and its implementation in the surface protocol, which allows very efficient filtering of the created objects. This implementation of the Machine learning workflow into Imaris is very appreciated.

Nicolas Goudin, Head of Necker Bioimage Analysis Platform – SFR Necker, Paris

Additional Resources

The Imaris Learning Center hosts a wide range of tutorial videos, how-to articles and webinars to guide you through the many features of Imaris. We have provided some links below which will get you started on some of our most recent developments.

Release Notes: Imaris 10.1
Brochure: Imaris
Tutorial: AI Machine Learning Segmentation in Imaris – Introduction and Training Tips
Tutorial: AI Machine Learning Segmentation for Multiple Images – Re-Training the Classifier and Segmentation in a Batch Mode

このウエブサイトは、ミラマー・コミュニケーション Ltd