MVA Features

Deployment & User Interface

On-site / on-premise deployment

MVA is installed locally, on your infrastructure, inside your institute / hospital / company, giving you full control over your data and annotations.

Trivial single file deployment and installation on Linux and macOS, with very low hardware requirements

Cloud deployment

Alternatively, MVA can be hosted in the cloud, on infrastruture operated and monitored by MVA

User Interface

The user interface, based on WebAssembly (WASM), runs in all modern web browsers and allows to use MVA from anywhere (including tablets) – no installation necessary

The system is designed to provide a fluid user experience for 3D annotations even on commodity hardware and slow / high-latency internet connections, e.g. for world-wide research co-operations

Custom Adaptations

Adaptation to your requirements

Every annotation project is different. Designed to be flexible, MVA can be adapted to suit your needs. You can for example specify individual workflows, data formats or reporting requirements

Development of new features

Going beyond the adaptation of existing functionality, entirely new features can be developed for your project, in tight collaboration between MVA and your institution.

For example optimized editing tools and in-browser visualizations or integration with your in-house user & data management and existing AI infrastructure

Import of existing annotation projects

Large scale research projects run for years. MVA makes it possible to import these results and continue the annotation project with MVA.

Additionally, there is no lock-in – you can always export all your data from MVA

Annotation

2D / 3D viewer

Fast in-browser viewer and editor, for any 2D and 3D image type, up to whole-body CTs. Support for switching between imaging sequences for MRI and support for Hounsfield units incl windowing presets

Classic / superpixel based brushes

Annotate your 2D and 3D images with brushes or superpixel based tools

Multiple labels per voxel

Assign a single or multple labels per voxel, e.g. for anatomy and pathology, or for overlapping pathological areas

Propagation between slices

Propagate the labels from one slice to the next, automatically deforming the annotations to match the new image

Slice and volume labels

Instead of labelling individual pixels / voxels, assign labels to entire slices or the whole volume

Negative slice and volume labels

Specify that a label is absent in a slice or volume. Mutual exclusivity between positive and negative labels is ensured throughout voxels / slices / volume

Edit existing annotations

Upload existing annotations or model predictions together with the image data - correct them in MVA or get expert feedback

Label hierarchies

Use multiple, hierarchically organized label sets, for example for anatomical structure and pathologies. Map the labels to semantic ontologies like Radlex and Snomed

Keyboard driven

Keyboard hotkeys for fast, strain-free editing – no need to hold the mouse button all day long

Rectangles / Circles / Polygons

Draw bounding boxes or more complicated shapes quickly

coming soon

AI brushes and auto-segmentation

AI-powered brushes and labelling tools based on state of the art models

on roadmap

Workflows

Projects / Workflows

A project in MVA models the propagation of the images through the steps of import / annotation / review / export. Projects can have arbitrary structure, with annotations and review steps mixed at will. Each step can have different users assigned with specific settings what each user is allowed to do at each step.

For example, a quick initial screening node with simple "Yes / No" input can be followed by multiple rounds of detailed voxel annotation and review steps. Cases can be reassigned and even the workflow can be reconfigured while an annotation campaign is running

Custom Forms

Custom forms can be designed inside MVA, with checkboxes / radiobuttons / text fields. These can be used per slice and per volume to ask the annotators or reviewers complex questions about the data. Input in these forms can also be configured to affect the progression of a case through the workflow

Users / Teams / Roles / Permissions

MVA allows fine-grained control over users / teams and their roles and permissions. Control who sees what, who can annotate / review cases and who can manage the system, with preconfigured roles for ML devs, annotators, reviewers and project managers

Dashboard / Statistics / Reporting

For each project fine grained status information is avaible on overall progress and individual cases' status. See user and team time tracking information, productivity and review scores on a daily / weekly / monthly basis. All information can be visualized inside MVA and exported as highly detailed Excel documents

Custom workflow steps

As part of the customization pledge by MVA, new workflow steps can be implemented which map exactly to your needs, may it be annotation / review tasks, interaction with your infrastructure or notification channels

Data Import / Export

Data transfer

Image data including existing annotations can be imported and exported either via a command line utility or via Amazon S3 (or other S3 compatible) data buckets

Image Import

Image data can be imported as NIfTI volumes or DICOM series, as well as common 2D image formats

Annotations Import

Existing annotations can be imported along the image data as NIfTI volumes

Annotations Export

Pixel / voxel annotations are exported as NIfTI volumes. Positive as well as negative slice and volume labels are exported as JSON