Compare your options
Compare specialist editors, data platforms and labelling services against your CT or MRI workflow.
Real clinical CT with four radiologists' outlines, Armato et al., "Data From LIDC-IDRI", TCIA, CC BY 3.0 (opens in a new tab), modified. Dashed: model proposal from SAM 2.1, outside Nitsor. Sources
Team costs
No per-seat fees.
Add annotators and reviewers without paying for each person.
CVAT Online Team bills per person on its monthly and yearly plans. Supervisely Pro lists fees for additional members and for labeling members.
Nitsor has no per-person charge on Community, Team or Enterprise. Workspace subscriptions, model-assisted volume usage and fair use follow each Nitsor plan.
Paid-plan collaboration terms checked 7 October 2026: CVAT Online Team pricing (opens in a new tab), Supervisely Pro pricing (opens in a new tab).
See Nitsor plans and usageSoftware or services
Decide who does the labelling.
A service supplies the people who label. Software gives your own team the place to do it. Both are good businesses, and plenty of teams use both.
- A labelling service
- A provider supplies annotators and delivers work to an agreed scope. Review responsibilities, data handling and handover belong in the contract.
- Your team's workspace
- Your team labels and reviews scans in Nitsor. Existing scans remain in connected storage; uploads use your selected destination. The browser reads viewed slices through short-lived signed links, and model jobs use temporary copies while they run.
Find the right fit
Choose around your team's responsibilities.
Choose a service when
- You need annotation capacity outside your team.
- You have a defined dataset and delivery scope.
- The provider's data handling meets your requirements.
- You need the provider to supply qualified reviewers.
Choose Nitsor when
- Your experts inspect and revise model proposals.
- You want to connect existing object storage.
- Your team revises labels and class definitions over time.
- You need model suggestions and human decisions identified separately.
You can also use both. A provider's annotators can work in your Nitsor project under named accounts with the role you grant, and the record names them the way it names your own people.
Question by question
What to ask any provider, and how Nitsor answers.
| Question | Ask a service or platform provider | Nitsor |
|---|---|---|
| Where are scans stored? | Which storage locations and processing copies does the workflow use? | Connect supported object storage. The browser reads viewed slices from it through short-lived signed links, and hosted model jobs use temporary copies while they run. |
| What accompanies a label? | Can we retrieve authorship, revisions and review decisions? | Saved label revisions keep their author and the reviews recorded against them. |
| Who sets review policy? | Which rules decide what gets reviewed, and is that decision recorded? | Route work by model score or sampling, with the routing decision recorded. |
| Who does the work? | Can our experts and outside contributors work in the roles we need? | Invite contributors and reviewers under named accounts, with no per-person charge. |
| What can an auditor inspect? | Can they inspect the working history as well as a report? | Give scoped read-only access to labels, reviews and dataset versions. |
| What can we retrieve when leaving? | Which files and records come with the handover? | Read labels and history through documented interfaces and plain exports. |
| What determines the bill? | Which charges are platform, labour, delivery and usage? | Choose Community, Team or Enterprise. No plan meters people, labels or reads. Team meters one thing: scans run through a hosted model beyond the 250 included in each monthly billing period. |
| Published model-run price | Is the price of a model run published, and do you see it before the run starts? | Yes. After the 250 volumes Team includes in each monthly billing period, one scan run through one model version costs $0.25 per extra volume. Each run shows its quote for confirmation before it starts. |
By category
Questions to ask in an evaluation.
These are questions for your evaluation, not claims that a product lacks a feature. Each vendor documents its own capabilities, and the links go to that documentation, checked 27 Sep 2026.
| Category | Examples, from their documentation | Question to ask | Documentation |
|---|---|---|---|
| General annotation platforms | CVATIts documented 3D mode draws cuboids around objects in a 3D scene. | Does the workflow handle volumetric masks, linked views and the review-to-training handoff we need? | |
| Specialist imaging annotation platforms | RedBrick AI, EncordRedBrick AI documents web annotation for CT, MRI and X-ray with project management and quality control. Encord documents DICOM annotation with current transfer syntaxes. | Can we follow a corrected label through review into the exact dataset used for training? | |
| Dataset exploration and evaluation | FiftyOneIts documentation covers exploring data, finding likely label mistakes, and annotating through integrations such as CVAT, Label Studio and Labelbox. | How does dataset exploration connect to annotation, expert review and saved training inputs? | |
| Data versioning tools | DVC, lakeFSDVC tracks large datasets and models alongside code in Git. lakeFS gives object storage Git-style branches, commits and merges. | How much annotation and review behaviour must our team build around versioning? | |
| Desktop imaging tools | 3D SlicerSoftware for visualising, segmenting, registering and analysing medical and other 3D images. | How will several people assign work, review revisions and keep a shared history? | |
| Managed labelling services | Labelling service providersScope, review and handover are set by each provider's contract. | Who supplies the experts, what is delivered, and which working records stay with us? | A category, no product named |
| Scripts, files and spreadsheets | Your existing workflowThe links are whatever your team maintains by hand. | Who maintains the links between scan revisions, masks, review decisions and training runs? | A category, no product named |
Free tools
Nitsor beside 3D Slicer, ITK-SNAP and CVAT.
Three free, open-source tools for 3D images and annotation. Each cell comes from the project's own documentation, checked 28 Sep 2026.
| Check | Nitsor | 3D Slicer | ITK-SNAP | CVAT |
|---|---|---|---|---|
| 3D volumes and DICOM | Opens uncompressed DICOM Part 10 CT and MRI series and VGStudio volumes in linked 3D views. | Opens CT and MRI volumes. Its DICOM module imports, browses and exports series. | Opens 3D and 4D volumes, with a DICOM series selector. | Opens images, video and 3D point clouds. DICOM is first converted to 2D images. |
| Review and second read | A second reader accepts, corrects or rejects each piece of work, and both names stay on record. | No review workflow documented. A segment can be flagged for an expert reviewer. | No review workflow documented. | Review mode: reviewers raise issues and annotators fix them. Not for 3D tasks. |
| Dataset versioning | Fixed dataset versions that training jobs pin, and a History of every revision with its author and change counts. | No dataset versioning documented. | No dataset versioning documented. | No dataset versioning documented. Tasks and projects can be backed up and exported. |
| Model-assisted labelling | SAM 2.1 proposals on Nitsor compute, kept apart until a person accepts, corrects or rejects each. Not in Community. | Through extensions such as MONAI Label and TotalSegmentator, plus semi-automatic tools in the core. | nnInteractive since version 4.4, through a separate deep learning server with a GPU. | Automatic annotation with models. Self-hosted instances run models their administrators deploy. |
| Hosting and self-hosting | Community is self-hosted. Team is hosted by Nitsor. Enterprise is hosted or privately deployed. | Desktop app for Windows, macOS and Linux. Also runs in containers. | Desktop app for Windows, macOS and Linux. | Web app. Self-host the free Community edition, or use the project's hosted edition. |
| Licence | Source-available, under FSL-1.1-ALv2. | BSD-style 3D Slicer License. | GNU GPL v3. | MIT for the Community edition. The Enterprise edition has a commercial licence. |
Versions checked: 3D Slicer 5.12.4, ITK-SNAP 4.4.0, CVAT 2.76.0. Sources: 3D Slicer documentation (opens in a new tab), ITK-SNAP documentation (opens in a new tab), CVAT documentation (opens in a new tab).
Before and after
Find the labels behind a model.
Some teams run a drawing tool, a shared drive and a spreadsheet. As datasets change, they need a reliable link between source scans, label revisions, review decisions and training inputs.
| Question | Drawing tool and shared drive | Nitsor |
|---|---|---|
| Where the labels are | Files named final, final_v2 and final_v3_use_this. | Submitted labels and finished reviews are commits, each with a parent, so the chain from first label to pinned version is one list. |
| Who drew it | Whoever had the drive mapped. The spreadsheet says, if somebody filled it in. | The record says, because the name is part of the commit rather than a column beside it. |
| What changed | Compare the files and reconcile the changes. | Every revision in History shows what it added, removed and changed. |
| Which labels trained the model | Reconstruct the file list and revisions the training job read. | One dataset version ID and its manifest hash. |
| What it costs | Software, infrastructure and maintenance effort. | Community is free to self-host. Team is $649 per workspace per month or $6,228 upfront per workspace per year ($519 per month equivalent), with 250 model-assisted volumes per monthly billing cycle, then $0.25 per extra volume. Enterprise starts at $12,000 per year, plus your volume commitment. These are launch prices. Paid checkout and usage charging are not active. |
Show us what you inspect.
Start with a public CT or MRI scan. See how a model suggestion becomes a reviewed label and a fixed training dataset.