Best AI Auto-Contouring Software 2026: Vendor Comparison

Radformation AutoContour

Best for: High-volume clinics in the Varian/Eclipse ecosystem that want the largest model library plus planning-structure automation.

Radformation AutoContour (v2.7) is one of the most established auto-segmentation tools on the market, distinguished above all by the breadth of its model library — 480 guideline-based models, including 115 dedicated lymph-node models — across CT, MR, and CBCT. It sits inside the broader Radformation suite (ClearCheck, ClearCalc), and its tight Eclipse API integration makes it especially compelling for departments already standardized on Varian. As with virtually all tools in this category, it contours OARs, elective nodal CTVs, and planning structures, not tumor/GTV.

• 480 guideline-based models (115 lymph-node models) — the deepest library among mainstream vendors, with TG-263-compliant nomenclature.
• Modalities: CT, MR, and CBCT (useful for adaptive/IGRT workflows).
• Editor: pencil, adaptive brush, interpolation, and Edit Assist for fast manual correction (desktop/server install).
• Integration: DICOM RTSTRUCT to any TPS, plus a direct Eclipse API path for Varian shops.
• Deployment: on-prem or cloud, via desktop/server installation.
• Regulatory: broad clearances — FDA 510(k), EU MDR CE, TGA, ANVISA, Health Canada, and Thai FDA.

Pricing: Quote-only. Pricing is not published; expect site-license/per-clinic quoting, often bundled with the wider Radformation suite.

How it compares to AutoSeg. AutoContour and AutoSeg solve overlapping problems with different center of gravity. AutoContour's strength is sheer library depth (480 models), planning-structure automation, and a mature Eclipse API path — ideal if you live in the Varian world and value the largest catalog plus an integrated QA suite. AutoSeg's differentiators lie in workflow and openness: a 100%-in-browser, multi-platform editor with real-time collaborative editing (per-structure locks and live presence), an end-to-end zero-click pipeline that returns the RTSTRUCT all the way back to the TPS, and vendor-neutral DICOM integration with no scripting or plugins. AutoSeg also offers on-prem, cloud, and air-gapped deployment, plus a built-in review→approval→audit workflow (RBAC + WORM audit). On regulatory breadth AutoContour is clearly ahead today (FDA/CE and more), whereas AutoSeg holds ANVISA RDC 657/2022 and is built to ISO 13485 / IEC 62304 Class C without claiming FDA/CE. The honest framing: choose AutoContour for catalog depth and Varian-native automation; choose AutoSeg when browser-based collaboration, vendor-neutrality, and air-gap flexibility matter more.

When to choose it: You are a high-volume, Varian/Eclipse-centric department that wants the broadest model and nodal library, planning-structure automation, and the widest set of regulatory clearances.

When to avoid it: You need real-time multi-user contouring, a fully browser-based editor, vendor-neutral no-scripting integration across mixed TPS vendors, or an air-gapped deployment.

Verdict: The deepest model library and strongest planning automation in the Varian/Eclipse ecosystem, backed by the broadest regulatory clearances — but without real-time collaboration or a browser-based editor. A safe, capable default for Varian-heavy clinics.

Limbus AI (Limbus Contour)

Best for: Clinics wanting a proven, local, no-GPU standalone auto-contouring install with broad regulatory clearance

Limbus AI (Limbus Contour) is one of the most established names in deep-learning auto-contouring, with a deep base of independent peer-reviewed validation behind it. Its signature strength is operational simplicity: a standalone workstation install on Windows 10+ that runs without a dedicated GPU and keeps PHI local, making it unusually easy for IT-constrained departments to adopt. Following Radformation’s 2024 acquisition, Limbus Contour is being folded into the unified Radformation AutoContour platform, so buyers should clarify which product line and license they are actually purchasing.

• 260+ structures (OARs plus elective nodal CTVs) with TG-263 nomenclature support
• Modalities: CT, MR, and CBCT
• Standalone workstation install (Windows 10+) that runs without a dedicated GPU; on-prem local processing (no cloud PHI) plus a cloud option
• Built-in review/edit step to inspect and adjust contours before export
• Vendor-neutral DICOM RTSTRUCT output, with Eclipse (ESAPI) integration
• Broad regulatory clearance: FDA 510(k) K201232, EU MDR CE (2025), TGA, ANVISA, Health Canada, and Thai FDA

Pricing: Quote-only; no public pricing. Confirm whether you are licensing legacy Limbus Contour or the unified Radformation AutoContour, and on what per-seat or per-workstation basis.

How it compares to AutoSeg. Limbus and AutoSeg solve the same core problem but with different architectures. Limbus excels as a local, low-IT-footprint desktop install with deep published validation and broad multi-region clearance — strong reassurances for risk-averse procurement. AutoSeg (https://rtmedical.com.br/en/rtconnect/) differs in deployment and workflow: a 100% in-browser editor with real-time collaborative editing (per-structure locks and live presence), end-to-end zero-click automation including RTSTRUCT return to the TPS, vendor-neutral DICOM with no scripting or plugins, deployment across on-prem, cloud or air-gapped Kubernetes, and a built-in review→approval→audit (RBAC + WORM) governance workflow. On regulatory breadth Limbus is currently ahead — it holds FDA 510(k) and EU MDR CE marks, which AutoSeg does not claim (AutoSeg cites ANVISA RDC 657/2022 and is built to ISO 13485 / IEC 62304 Class C / ISO 14971). Buyers prioritizing a proven standalone install with the widest clearances may prefer Limbus; those prioritizing browser-based collaboration, zero-click TPS round-trip and flexible deployment will favor AutoSeg.

When to choose it: Choose Limbus when you want a battle-tested, locally installed auto-contouring tool with minimal IT overhead (no GPU, PHI on-site) and need broad regulatory clearance — especially FDA 510(k) or EU MDR CE — for procurement in your jurisdiction.

When to avoid it: Avoid it if you need real-time multi-user collaborative editing, a fully browser-based multiplatform editor, end-to-end zero-click RTSTRUCT delivery back to the TPS, or air-gapped Kubernetes deployment — or if the post-acquisition migration into AutoContour makes long-term product continuity a concern.

Verdict: A mature, well-validated and low-IT-barrier local auto-contouring solution with class-leading regulatory breadth; the main caveat is the ongoing consolidation into Radformation AutoContour, which buyers should clarify before signing.

MVision AI Contour+

Best for: Clinics that prioritize broad guideline-based nodal coverage plus a dedicated Dice/Hausdorff QA tool.

MVision AI Contour+ is a cloud-first auto-contouring service from Finland-based MVision AI, paired with the Workspace+ platform and the separate Verify QA module. Its strongest selling point is breadth of guideline-aligned coverage: 300+ ROIs, including roughly 90 lymph-node regions mapped to 25+ published clinical guidelines, which is genuinely useful for elective nodal CTV work. Like nearly all tools in this category, it segments OARs and elective nodal volumes rather than tumor/GTV.

• 300+ ROIs across CT and MR, including ~90 lymph-node areas aligned to 25+ clinical guidelines
• Verify QA tool computes Dice and Hausdorff distance of contours against a reference set
• Cloud service with a browser-based platform (Workspace+); on-prem deployment available on request
• DICOM integration with Eclipse (via API), RayStation and Monaco
• Strong regulatory clearances: FDA 510(k), EU MDR CE mark, Australia TGA, and others
• No integrated freehand editor: corrections are made downstream in the TPS

Pricing: Quote-only; pricing is not published and depends on site, volume and deployment model.

How it compares to AutoSeg. MVision and AutoSeg solve overlapping problems with different philosophies. MVision leads on sheer catalog breadth (300+ ROIs, ~90 nodal areas) and on its standalone Verify QA tool with Dice/Hausdorff metrics, plus a wider set of regulatory clearances including FDA 510(k) and EU MDR CE — areas where AutoSeg, cleared under ANVISA RDC 657/2022 and built to ISO 13485 / IEC 62304 Class C / ISO 14971 (no FDA/CE claimed), is more regional today. Where AutoSeg differentiates is the editing and review experience: a 100% in-browser, multiplatform editor with real-time collaborative editing (per-structure locks and live presence), versus MVision's model where corrections move back into the TPS with no in-app collaboration. AutoSeg is also end-to-end zero-click including RTSTRUCT return to the TPS, uses pure vendor-neutral DICOM with no scripting or plugins, ships on-prem/cloud/air-gapped from the same Kubernetes-native build, and carries a built-in review→approval→WORM-audit workflow with RBAC and SSO. For a clinic whose priority is the widest possible guideline-mapped nodal library plus quantitative QA, MVision is a strong fit; for one that wants collaborative in-browser editing and a closed-loop, auditable contouring workflow, AutoSeg fits better.

When to choose it: Choose MVision if your top priorities are maximal guideline-mapped nodal coverage, quantitative QA via the Verify Dice/Hausdorff tool, and broad international clearances (FDA 510(k), CE, TGA), and your team is comfortable editing contours inside the TPS.

When to avoid it: Avoid it if you need an integrated browser-based editor, real-time multi-user collaboration, a fully zero-click round-trip back to the TPS, or an on-prem/air-gapped-first deployment with a built-in approval-and-audit workflow.

Verdict: Excellent guideline-aligned nodal coverage backed by a genuine quantitative QA tool, but no integrated freehand editor and no collaboration — best as a high-breadth contour generator that hands editing back to the TPS.

MIM Contour ProtégéAI+

Best for: Clinics already standardized on the MIM platform that want true zero-click contouring triggered at CT simulation.

MIM Contour ProtégéAI+ (v2.0) is the deep-learning auto-contouring module inside MIM Software’s radiation-oncology platform, now part of GE HealthCare. It auto-segments roughly 159 structures across CT and MR, can fire automatically the moment a CT-SIM dataset arrives, and routes results back through DICOM-RT to your TPS. It is one of the few solutions in this guide carrying FDA 510(k) clearance, including for the 2026 v2.0 release, plus an FDA-accepted PCCP for shipping future models.

• ~159 contoured structures (OARs and nodal levels) across CT and MR — no tumor/GTV segmentation
• True zero-click workflow: auto-triggers at CT-SIM via MIM Workflows, no manual launch
• Review and editing inside the established MIM viewer with automated contour QA protocols (Dice/HD reported in validation)
• Vendor-neutral TPS reach over DICOM-RT: Eclipse, RayStation, Monaco, Pinnacle
• TG-263 compliant, customizable naming templates
• Deploys on-prem or cloud on the MIM desktop/server platform

Pricing: Quote-only; pricing is not published and depends on platform footprint, modules, and deployment.

How it compares to AutoSeg. Both tools deliver vendor-neutral DICOM-RT integration and zero-click automation, and MIM has a regulatory edge for US buyers with its FDA 510(k) clearance and PCCP — AutoSeg is cleared under ANVISA RDC 657/2022 and built to ISO 13485 / IEC 62304 Class C / ISO 14971 without claiming FDA or CE. Where they diverge is workflow architecture: AutoSeg's editor is 100% in-browser (Cornerstone3D/WebGL2) and runs on Windows, Mac, Linux, and tablets with no install, whereas MIM's review and editing happen in the desktop/server MIM viewer. AutoSeg adds real-time collaborative editing — per-structure locks and live presence, a "Google Docs of contouring" model — plus a structured contour_reviewer to contour_approver workflow with WORM audit, while MIM is a strong single-user review experience with automated QA protocols. Both push RTSTRUCT back to the TPS; AutoSeg's pure DICOM (C-STORE/C-ECHO) approach requires no scripting or plugins, and it explicitly supports on-prem, cloud, and air-gapped deployments. MIM is the safer choice if you are already invested in its platform and need FDA paperwork; AutoSeg is the choice if browser-based multi-user review and air-gapped flexibility are priorities.

When to choose it: Choose MIM if you already operate the MIM platform, need FDA 510(k)/PCCP regulatory coverage, and want auto-contouring that fires automatically at CT simulation.

When to avoid it: Avoid if you want a zero-install, browser-based editor with real-time multi-user collaboration, or if you are not prepared to adopt the broader MIM ecosystem.

Verdict: A mature, FDA-cleared, vendor-neutral auto-contouring solution that is hard to beat for clinics already running MIM. Less compelling if browser-based collaborative review or a lighter-weight footprint is what you are after.

Varian Ethos / AI-Rad Companion Organs RT

Best for: Varian/Siemens Healthineers departments standardized on Eclipse, ARIA, and Ethos online adaptive therapy

Varian Ethos / AI-Rad Companion Organs RT (Siemens Healthineers) packages AI auto-contouring across two tightly integrated paths: cloud-based AI-Rad Companion Organs RT for offline planning and the on-board Ethos engine that drives daily online adaptive radiotherapy on CBCT/HyperSight. It is purpose-built for shops already living inside the Varian ecosystem, where contours flow natively into Eclipse, Velocity, and ARIA without bolting on a third-party system.

The standout is Ethos adaptive: segmentation, influencer/target review, and re-planning happen at the console while the patient is on the table — a workflow few competitors match clinically. For conventional planning, AI-Rad Companion Organs RT covers a broad OAR set delivered through the teamplay cloud.

• AI-Rad Companion Organs RT: 108 OAR models (CT); Ethos engine: 70+ structures for adaptive workflows
• Modalities limited to CT and CBCT — no MR-cleared OAR models
• Auto-segments OARs only; targets (GTV/CTV) remain research/manual, not a cleared feature
• FDA 510(k)-cleared (AI-Rad Organs RT, Ethos 2.0) and CE-marked — strong regulatory posture for US/EU sites
• No standalone contouring editor — editing happens in Eclipse, Velocity, or the Ethos console
• Deployment as cloud (AI-Rad via teamplay) plus on-prem (Ethos/Velocity)

Pricing: Quote-only, typically bundled into a teamplay cloud subscription and/or the Ethos/Eclipse licensing structure. No public list pricing; expect ecosystem-level commercial terms.

How it compares to AutoSeg. Both tools auto-segment OARs (neither contours tumor/GTV as a cleared feature), but they target different buyers. Varian's strength is depth inside its own ecosystem — Ethos online adaptive and native Eclipse/ARIA flow — paired with FDA/CE clearances AutoSeg does not currently claim. AutoSeg's differentiators are orthogonal: a 100% in-browser, multiplatform editor with real-time collaborative editing (per-structure locks plus live presence), versus Varian's edit-in-the-TPS model with no collaboration. AutoSeg is vendor-neutral via pure DICOM (C-STORE/C-ECHO + RTSTRUCT) with no scripting or plugins — connecting to Eclipse, Monaco, RayStation and others — and offers end-to-end zero-click automation including RTSTRUCT return to the TPS, where Varian keeps you within its stack. AutoSeg also adds user-selectable multi-model AI per protocol, CT and MR coverage, on-prem/cloud/air-gapped deployment, and an RBAC review to approval to WORM-audit workflow. If your site is all-Varian and prioritizes adaptive and cleared regulatory status, Varian is hard to beat; if you need cross-vendor neutrality, collaborative browser-based editing, MR, or air-gapped deployment, AutoSeg fits better.

When to choose it: Choose Varian if your department is standardized on Eclipse/ARIA and especially if you run or plan Ethos online adaptive therapy on CBCT/HyperSight, and FDA/CE clearance is a procurement requirement.

When to avoid it: Avoid if you run a mixed-vendor TPS fleet, need MR-based OAR segmentation, want a vendor-neutral browser editor with real-time collaboration, or require air-gapped deployment outside the Varian/teamplay stack.

Verdict: The strongest choice inside the Varian ecosystem — outstanding for Ethos adaptive and native Eclipse/ARIA integration — but CT/CBCT-only, with no MR-cleared OARs, no standalone editor, and no real-time collaboration.

RayStation Deep Learning Segmentation

Best for: RayStation TPS sites wanting native, in-TPS deep-learning contouring with fully local inference.

RayStation Deep Learning Segmentation (RaySearch Laboratories) is auto-contouring built directly into the RayStation treatment planning system rather than a standalone product. The v2025 release ships 201 deep-learning segmentation models spanning CT, CBCT and MR, with inference running on a local GPU inside your existing RayStation install. For departments already standardized on RayStation, it is one of the most tightly integrated options on the market — contours land natively in the planning workspace with no separate application, export step or DICOM hop.

• 201 deep-learning segmentation models (v2025), focused on organs at risk and elective nodal CTV — not tumor/GTV.
• Modalities: CT, CBCT and MR, broadening utility for adaptive and MR-guided workflows.
• Local GPU inference inside the RayStation TPS — no data leaves the site, no cloud dependency.
• Native RayStation editing tools (smart brush, interpolation) for review and correction in the same workspace.
• DICOM, RayCare OIS integration, and Eclipse interoperability via scripting.
• Mature regulatory pedigree: FDA 510(k) for the RayStation ML segmentation (8B); CE/MDR. Note the newest v2025 DLS is “subject to clearance in some markets.”

Pricing: Quote-only. Delivered as a feature/option of the RayStation TPS rather than a separately priced product; pricing is negotiated as part of your RayStation licensing.

How it compares to AutoSeg. The two solve different problems. RayStation DLS is the right answer when RayStation is your TPS: native, local, private, and integrated, with structures appearing directly in the planning workspace. AutoSeg is vendor-neutral — it connects to any TPS (Eclipse, Monaco, RayStation, and others) over pure DICOM (C-STORE/C-ECHO + RTSTRUCT) with no scripting or plugins, and returns the RTSTRUCT end-to-end zero-click. Editing and review also differ: AutoSeg runs a 100% in-browser editor with real-time multi-user collaboration (per-structure locks plus live presence) and a multi-tenant review→approval→WORM-audit workflow, whereas RayStation editing is single-user inside the TPS. AutoSeg additionally offers on-prem, cloud and air-gapped deployments and user-selectable multi-model AI per protocol, while RayStation is on-prem and uses its own model set. If you are committed to RayStation and want everything inside the TPS, RayStation DLS is excellent; if you run a mixed or multi-vendor estate, or want collaborative browser-based review with a formal approval trail, AutoSeg fits better.

When to choose it: Choose it if RayStation is your planning system and you want auto-contouring that lives natively in the TPS with fully local inference and no external data flow.

When to avoid it: Avoid it if you do not run RayStation, need cloud or air-gapped deployment, want vendor-neutral DICOM integration across multiple TPSs, or require real-time collaborative editing and a formal review/approval/audit workflow.

Verdict: Excellent if RayStation is your TPS — native, local, private and well-integrated; largely irrelevant if it is not. Remember the model count reflects models, not distinct anatomies, and it targets OARs/nodal CTV rather than tumor.

Mirada DLCExpert

Best for: Clinics that want the original FDA-cleared deep-learning contour engine with hands-off server automation.

Mirada DLCExpert from Mirada Medical was the first deep-learning auto-contouring product to earn FDA 510(k) clearance (2018), and it remains a credible, clinically proven OAR engine. It runs as a server-side “Workflow Box” (with an optional DLCOnline cloud path) that ingests CT, generates contours from consensus-guideline-trained models, and pushes results downstream with minimal user interaction. It is a focused, mature tool rather than a broad multi-modality platform.

• FDA 510(k)-cleared (2018, first DL auto-contouring) and CE marked
• ~99 CT organ-at-risk structures from consensus-guideline-based models
• Zero-click, server-side automation via the Workflow Box; DLCOnline cloud option
• CT-only; no MR auto-contouring
• Standard DICOM in/out; TPS interoperability via export (no named deep connectors)
• External validation reported as Dice/Hausdorff metrics; no built-in QA dashboard

Pricing: Quote-only; pricing depends on deployment (Workflow Box vs. DLCOnline) and structure scope. Request a formal quote from Mirada.

How it compares to AutoSeg. Mirada and AutoSeg share a zero-click automation philosophy, and Mirada has the stronger regulatory pedigree with its FDA 510(k) and CE marking (AutoSeg is ANVISA RDC 657/2022 cleared and built to ISO 13485 / IEC 62304 Class C, but does not claim FDA/CE). Where they diverge is workflow and scope. AutoSeg is multi-modality (CT and MR, including dedicated brain, male-pelvis, gyn, and neural nnU-Net models) and bundles a 100% in-browser editor with real-time multi-user collaboration — per-structure locks and live presence — so review and correction happen in the same place the contours are generated. Mirada is CT-only and pushes editing out to Mirada RTx or the TPS. AutoSeg's zero-click loop is end-to-end including RTSTRUCT return to the TPS, with vendor-neutral, no-scripting DICOM (C-STORE/C-ECHO) to any TPS, plus an explicit review-to-approval-to-WORM-audit RBAC workflow and on-prem/cloud/air-gapped deployment. If your need is a single proven CT OAR engine with the broadest clearances, Mirada is a strong, honest choice; if you need MR coverage and a collaborative in-browser approval workflow, AutoSeg fits better.

When to choose it: Choose Mirada if you primarily need CT OAR auto-contouring, value the most established FDA/CE regulatory footprint, and are comfortable doing edits in your TPS or Mirada RTx.

When to avoid it: Avoid it if you need MR auto-contouring, an integrated collaborative editor, a built-in approval/audit workflow, or zero-click RTSTRUCT return without site-specific export scripting.

Verdict: A pioneering, regulatory-mature deep-learning OAR engine with solid server automation, held back by being CT-only and keeping editing outside the tool. Best as a dependable CT contour engine rather than an all-in-one platform.

TheraPanacea ART-Plan (Annotate)

Best for: Clinics needing the widest modality and structure breadth, including regulatory-cleared tumor segmentation

TheraPanacea ART-Plan (Annotate) is a cloud-based auto-contouring platform from the Paris-based vendor TheraPanacea, positioned around the broadest modality and anatomy coverage in this guide. It spans CT, MR, CBCT and PET, and is one of the very few products whose regulatory clearances extend beyond OARs and nodal CTVs into selected tumor targets. For departments that want one cloud tool to cover diverse disease sites and imaging modalities, Annotate is a serious contender.

• 200+ structures (number of models, not distinct anatomies): OARs, lymph nodes, and some target volumes
• Four modalities: CT, MR, CBCT and PET — the widest imaging breadth in this comparison
• Cleared tumor segmentation: FDA 510(k) (including K242822 for tumors, 2025) covers select targets such as prostate CTVn and gyn CTVt
• Regulatory: FDA 510(k), CE Class IIb (MDR), and TGA — strong for US, EU and Australian markets
• Cloud web app with an integrated viewer/editor; DICOM export to any TPS
• Structure naming follows consensus guidelines (TG-263 conformance not explicitly stated)

Pricing: Quote-only; pricing is not published and depends on modality mix, structure sets, and volume. Request a quote directly from TheraPanacea.

How it compares to AutoSeg. The two products optimize for different things. TheraPanacea's edge is breadth and clearances: four modalities (adding CBCT and PET beyond AutoSeg's CT and MR) and FDA/CE/TGA clearances that uniquely extend to selected tumor targets — something AutoSeg does not claim, as AutoSeg carries ANVISA RDC 657/2022 and is built to ISO 13485 / IEC 62304 Class C without FDA 510(k) or CE-MDR. AutoSeg's differentiators lie in workflow and deployment: real-time collaborative editing in a 100% in-browser editor (per-structure locks plus live presence, the "Google Docs of contouring"), end-to-end zero-click automation including RTSTRUCT return to the TPS, vendor-neutral pure-DICOM integration with no scripting or plugins, flexible on-prem / cloud / air-gapped install, and a built-in reviewer to approver workflow with WORM audit. Annotate is cloud-only with a single-user cloud editor and DICOM export rather than a deep zero-click return loop. If cleared tumor segmentation and maximum modality breadth are decisive, TheraPanacea leads; if collaborative editing, deployment flexibility, and an integrated review-and-approval audit trail matter more, AutoSeg leads.

When to choose it: Choose TheraPanacea when you need CBCT or PET coverage, the largest structure library, or regulatory-cleared tumor/target segmentation, and a cloud deployment is acceptable.

When to avoid it: Avoid it if you require on-prem or air-gapped hosting, real-time collaborative contouring, or a fully zero-click RTSTRUCT return to your TPS.

Verdict: A standout for modality breadth and the rare cleared tumor segmentation, backed by FDA, CE and TGA. The trade-offs are its cloud-only model and lack of real-time collaboration or a zero-click TPS return loop.

Manteia AccuContour

Best for: Clinics wanting an all-in-one contouring, fusion, and dose-review workstation with self-training model support

Manteia AccuContour from Manteia Technologies is a feature-dense radiotherapy workstation that bundles AI auto-contouring with multimodal image fusion, a full viewer/editor, and dose-review tooling. It supports CT, MR, PET-CT, and 4D datasets, and pairs with a separate QA Solution and the optional Mozi TPS, positioning it as a near-complete contouring-to-planning bench rather than a single-purpose segmentation engine.

Marketing cites 300+ OARs and 20+ targets, but the FDA 510(k) clearance (K191928, 2020) is scoped to CT OAR contouring only — a distinction clinical buyers should verify against the structures they actually intend to use.

• Models/structures: 300+ OARs and 20+ targets marketed; cleared scope is CT OAR-only (note: “structure count” reflects models, not distinct anatomies)
• Modalities: CT, MR, PET-CT, and 4D, with multimodal fusion for review
• Editor: Full desktop workstation viewer/editor for contour creation and correction
• Self-training: Supports site-specific model training to adapt to local protocols
• Integration: DICOM-RT export to any TPS; optional tight coupling with Manteia’s Mozi TPS
• Deployment: On-prem workstation, cloud SaaS, or hybrid; separate QA Solution with Dice/HD metrics reported in studies

Pricing: Quote-only; pricing varies by modules (contouring, fusion, QA Solution, optional Mozi TPS) and on-prem vs. cloud/hybrid deployment.

How it compares to AutoSeg. Both deliver auto-contouring with full-featured editors, but the workflow philosophies differ. AccuContour is a powerful workstation-centric ecosystem — fusion, dose review, QA, and an optional in-house TPS — best suited to sites that want one vendor's bench. AutoSeg is built around a 100% in-browser editor (Cornerstone3D/WebGL2) with real-time collaborative editing — per-structure locks and live presence — that AccuContour's installed-workstation model does not match. AutoSeg also emphasizes end-to-end zero-click automation including RTSTRUCT return to the TPS, vendor-neutral DICOM integration (C-STORE/C-ECHO) with no scripting or plugins against any TPS, and on-prem/cloud/air-gapped Kubernetes deployment, plus a formal physician→reviewer→approver RBAC workflow with WORM audit. On regulatory pathway, AccuContour holds an FDA 510(k) (CT OAR) where AutoSeg does not claim FDA/CE; AutoSeg is built to ISO 13485 / IEC 62304 Class C / ISO 14971 under ANVISA RDC 657/2022. The honest choice hinges on whether you want an integrated single-vendor workstation suite (AccuContour) or a collaborative, vendor-neutral, browser-based contouring pipeline (AutoSeg).

When to choose it: Choose AccuContour if you want a single-vendor workstation that combines auto-contouring, multimodal fusion, dose review, and QA — and value an FDA-cleared CT OAR pathway plus the option to self-train models or add the Mozi TPS.

When to avoid it: Avoid it if you need real-time multi-user collaborative editing, a fully browser-based multiplatform editor, strictly vendor-neutral no-scripting integration, or air-gapped Kubernetes deployment — and if marketed-vs-cleared scope ambiguity is a procurement concern.

Verdict: A feature-rich, multimodal contouring-and-fusion workstation with fast editing and a real FDA-cleared CT OAR pathway; just confirm that the structures you need fall within the validated scope rather than the broader marketed claims.

Carina INTContour

Best for: Air-gapped clinics wanting 100% on-prem, browser-based contouring with built-in contour QA metrics.

Carina INTContour (Carina Medical) is a browser-based auto-contouring application deployed entirely on-premise, with no cloud dependency, making it a natural fit for data-sovereignty-conscious and air-gapped radiotherapy departments. It segments 120+ structures (OARs plus nodal groups) across CT, MR, and PET, and is notable for shipping a built-in Contour QA Tool that reports DSC, Hausdorff Distance, Mean Surface Distance, and Surface-DSC, which few peers offer natively. It is one of the few vendors here cleared via FDA 510(k) (K212274, 2022).

• 120+ structures (OARs and nodal groups) auto-segmented across CT, MR, and PET.
• 100% on-premise, browser-based application — no cloud required, suitable for air-gapped networks.
• Built-in Contour QA Tool reporting DSC, Hausdorff Distance, Mean Surface Distance, and Surface-DSC.
• Browser viewer with smart-interpolation editing tools.
• FDA 510(k) cleared (K212274, 2022).
• DICOM connectivity plus Eclipse integration via ESAPI; runs as a standalone web app.

Pricing: Quote-only; pricing is not published and depends on deployment and structure-model scope.

How it compares to AutoSeg. Carina INTContour is the closest peer to AutoSeg on the two dimensions buyers care most about here: browser-based access and on-prem data sovereignty, and it goes further than many with its native QA-metrics tool and an FDA 510(k) clearance that AutoSeg does not claim. The differences are mostly about workflow scope. AutoSeg adds real-time collaborative editing in the browser (per-structure locks and live presence) where INTContour is single-user; AutoSeg delivers end-to-end zero-click automation including RTSTRUCT return to the TPS via autoseg-desktop; and AutoSeg's DICOM integration is vendor-neutral with no scripting or plugins (C-STORE/C-ECHO + RTSTRUCT to any TPS), whereas INTContour's Eclipse path uses ESAPI. AutoSeg can also run on-prem, cloud, or air-gapped rather than on-prem only, and ships a multi-tenant review-to-approval-to-WORM-audit workflow with RBAC and SSO. INTContour is a strong, regulatory-credentialed choice for a single-site air-gapped clinic; AutoSeg fits teams that need multi-user collaboration, flexible deployment, and a governed approval audit trail.

When to choose it: Choose INTContour if you need a 100% on-prem or air-gapped browser tool, value an FDA 510(k) clearance, and want built-in DSC/HD/MSD/Surface-DSC QA metrics for single-user contour review.

When to avoid it: Avoid if you need real-time multi-user collaboration, cloud or hybrid deployment flexibility, vendor-neutral no-scripting routing to a wide TPS mix, or a formal multi-tenant review-approval-audit workflow.

Verdict: A credible, regulatory-backed peer on browser access, data sovereignty, and built-in QA — best for air-gapped single-site clinics, but on-prem-only and without real-time collaboration.

Siemens AI-Rad Companion Organs RT / syngo.via RT

Best for: Siemens imaging sites wanting at-scanner (DirectORGANS) or cloud OAR contouring at scale

Siemens AI-Rad Companion Organs RT / syngo.via RT is the radiotherapy auto-contouring portfolio from Siemens Healthineers, spanning a cloud service (teamplay), the syngo.via RT workstation, and DirectORGANS contouring generated directly at SOMATOM CT scanners. It targets departments already invested in Siemens imaging that want OAR segmentation available from acquisition through to the TPS.

The headline coverage is 200+ structures across CT and MR, including brain metastases on MR. As with every vendor, treat the count as the number of validated models rather than distinct anatomies, and note the tool contours organs at risk and nodal volumes, not the tumor/GTV.

• 200+ contouring models across CT and MR, including brain mets on MR
• Three delivery modes: cloud (teamplay), syngo.via RT workstation, and DirectORGANS at-scanner on SOMATOM systems
• OAR and elective nodal volumes — no tumor/GTV auto-segmentation
• DICOM connectivity with documented integrations to Eclipse and Velocity
• FDA 510(k) cleared (2020) and CE marked
• Editing in syngo.via RT; the cloud product pushes results to the TPS for downstream review

Pricing: Quote-only, typically via a teamplay subscription. Pricing varies with deployment mode (cloud vs. workstation vs. at-scanner) and site scale; request a formal quote.

How it compares to AutoSeg. Both are credible OAR auto-contouring platforms, and Siemens brings clearances AutoSeg does not claim (FDA 510(k) and CE). The contrast is one of philosophy. Siemens is strongest as an integrated extension of its own imaging chain — DirectORGANS at the SOMATOM scanner is a real differentiator no browser tool replicates. AutoSeg, by contrast, is deliberately vendor-neutral and scanner-agnostic: pure DICOM (C-STORE/C-ECHO + RTSTRUCT) with no scripting or plugins, so it slots into Eclipse, Monaco, RayStation and mixed fleets without favoring one OEM. Editing is the clearest divergence — AutoSeg offers 100% in-browser, multi-user editing with per-structure locks and live presence ('Google Docs of contouring') plus a built-in physician → reviewer → approver workflow with WORM audit, whereas Siemens editing happens in the syngo.via RT workstation without real-time collaboration. AutoSeg also delivers end-to-end zero-click including RTSTRUCT return to the TPS, and supports on-prem, cloud, and fully air-gapped deployment for sites that cannot use a vendor cloud.

When to choose it: Choose Siemens if your department runs SOMATOM scanners and Siemens imaging infrastructure, wants at-scanner DirectORGANS contouring, and values having FDA 510(k)/CE-cleared models across a large multi-site cloud deployment.

When to avoid it: Avoid it if you need real-time collaborative editing, run a vendor-neutral mixed scanner/TPS fleet, require fully air-gapped deployment, or want a built-in review-and-approval audit workflow rather than workstation-based editing.

Verdict: A compelling, well-credentialed choice for Siemens scanner sites and large multi-site cloud rollouts. The lack of real-time collaboration and its ecosystem tilt are the main trade-offs for vendor-neutral buyers.

SegmentaR (Silva Ray)

Best for: Brazilian clinics wanting a national, ANVISA-registered, no-local-GPU OAR tool

SegmentaR (Silva Ray) is, to our knowledge, the first Brazilian-developed auto-segmentation tool to carry a domestic ANVISA registration, which makes it notable for clinics that prioritize a national supplier and local regulatory standing. Its scope is deliberately narrow: CT-only organ-at-risk (OAR) segmentation across three anatomical regions (head & neck, thorax, pelvis), delivered through a thin local client that sends studies to cloud GPU inference. There is no separate contouring editor and no tumor/GTV segmentation; structures are reviewed and edited inside your existing TPS.

• ~56 OAR models spanning head & neck, thorax, and pelvis (CT only)
• Cloud GPU inference via a thin local client — no on-site GPU hardware required
• DICOM-based exchange (no named TPS connectors advertised)
• ANVISA registration 83339490001; LGPD-aligned data handling (no FDA 510(k) or CE-MDR claimed)
• No published Dice/Hausdorff metrics and no independent validation located
• Editing performed in the TPS — no dedicated in-product editor advertised

Pricing: Quote-only; no public pricing published.

How it compares to AutoSeg. SegmentaR and AutoSeg overlap on automated CT OAR segmentation and a DICOM-based workflow, but their footprints differ markedly. SegmentaR is a focused, cloud-inference CT OAR tool with editing delegated to the TPS, whereas AutoSeg adds a 100%-in-browser editor with real-time multi-user collaboration (per-structure locks and live presence), end-to-end zero-click automation that returns the RTSTRUCT all the way back to the TPS, and a vendor-neutral, no-scripting DICOM integration that works with any C-STORE/C-ECHO-capable TPS. AutoSeg also offers on-prem, cloud, and air-gapped deployment plus a built-in review to approval to WORM-audit workflow with RBAC and SSO. Both carry ANVISA standing (neither claims FDA/CE). For a Brazilian clinic that only needs CT OAR coverage for the common regions and is comfortable editing in its TPS, SegmentaR is a credible, lightweight national option; teams needing broader modality/structure coverage, in-product editing, collaboration, or flexible deployment will find AutoSeg the wider fit.

When to choose it: Choose SegmentaR if you are a Brazilian clinic that wants a national, ANVISA-registered supplier, needs only CT OAR segmentation for head & neck, thorax, and pelvis, prefers cloud inference with no local GPU, and is content to review and edit contours inside your existing TPS.

When to avoid it: Avoid it if you need MR support, broader/structure-rich coverage, tumor or nodal CTV segmentation, an integrated editor, multi-user collaboration, on-prem or air-gapped deployment, or published/independently validated accuracy data before buying.

Verdict: A pioneering, ANVISA-registered Brazilian CT OAR tool that is genuinely useful for clinics wanting a national, GPU-free option — but its narrow scope, absence of an editor or collaboration, and lack of published accuracy data limit it to a focused OAR-acceleration role rather than an end-to-end contouring platform.