What it does. Deep-learning auto-contouring of OARs and CTVs plus auto-generation of optimization/planning structures from clinic templates, integrated with Radformation’s QA suite (ClearCheck/ClearCalc). Radformation acquired Limbus AI in 2024.

When to choose it: Eclipse-centric or multi-modality clinics wanting integrated automation plus QA.

When to avoid it: if you need a minimal single-purpose tool, or strictly on-premise with no cloud component.

What it does. Pre-trained, expert-curated models contour OARs and CTVs from CT/MR locally and export DICOM RTSTRUCT for review in the TPS. Now part of Radformation, with capabilities converging.

When to choose it: sites that mandate data never leave the local network, or with limited GPU hardware.

When to avoid it: if you want a cloud-scalable service or a single integrated automation + QA suite.

What it does. Deep-learning auto-contouring of OARs on planning images (head & neck, thorax, abdomen, pelvis), delivered through the Siemens Healthineers teamplay cloud platform as part of the AI-Rad Companion family.

When to choose it: existing Siemens Healthineers / teamplay customers wanting cloud OAR contouring inside that ecosystem.

When to avoid it: if you need CBCT/PET contouring, an offline tool, or a documented large editable structure library today.

Context — Varian Ethos (separate product). Varian Ethos is an online adaptive radiotherapy treatment platform that re-images and re-plans on the linac, with its own integrated AI/deformable contouring and an automated planning engine. Varian is a Siemens Healthineers company, but Ethos and AI-Rad Companion Organs RT are different products, marketed separately — Ethos is not part of this auto-contouring software comparison.

What it does. Deep-learning auto-segmentation built into the RayStation TPS. Guideline-based models run locally on a GPU; generated ROIs are edited with RayStation’s own structure-definition tools.

When to choose it: departments on RayStation wanting in-workflow contouring across modalities and treatment types.

When to avoid it: if you run a different TPS and don’t want to adopt RayStation just for contouring, or need fully hands-off segmentation.

What it does. Cloud (or local) AI auto-segmentation of OARs and lymph-node regions, with “guideline-based segmentation” trained to follow published consensus guidelines (ESTRO, EPTN, RTOG, UK SABR) to reduce inter-observer variability.

When to choose it: teams wanting guideline-standardized OAR/nodal contouring with vendor-managed updates, on Eclipse/RayStation/Monaco.

When to avoid it: if you require CBCT/PET contouring, confirmed local availability today, or a single tool that both generates and fully edits contours.

What it does. Deep-learning auto-contouring inside MIM’s vendor-neutral imaging platform. ProtégéAI+ adds AI-assisted editing/refinement on top of auto-generation. MIM is now part of GE HealthCare.

When to choose it: teams wanting vendor-neutral auto-contouring plus AI-assisted editing, decoupled from any single TPS.

When to avoid it: if you want contouring native to your existing TPS with no extra platform.

What it does. ART-Plan’s Annotate module auto-contours a large library of OARs, lymph nodes and CTV models, with its own web editor and mandatory qualified-user validation. The wider platform adds image fusion and synthetic-CT from MR/CBCT.

When to choose it: sites wanting a high-volume guideline-aligned engine plus synthetic-CT and fusion, on Eclipse-ARIA, RayStation or Monaco-Mosaiq.

When to avoid it: if you require a fully on-prem/air-gapped TPS-native plugin, or confirmed local availability before purchase.

What it does. Deep-learning OAR contouring delivered through Mirada’s Workflow Box “zero-click” appliance, which intercepts incoming CT studies, generates contours and auto-routes the RTSTRUCT to a destination. One of the first commercially available AI auto-contouring solutions.

When to choose it: high-volume automated OAR contouring on planning CT in a multi-TPS environment.

When to avoid it: when you need a built-in editor or confirmed CBCT/PET support.

What it does. Software-only deep-learning auto-segmentation producing an initial OAR/target set for review and editing, with in-browser tools, an automated DICOM workflow and Eclipse ESAPI integration. The vendor also offers on-site model retraining and geometric QA.

When to choose it: US (especially Eclipse) sites wanting a tool with local retraining and clinician review.

When to avoid it: when you need native RayStation integration, or confirmed availability in your region without vendor confirmation.

What it does. A standalone AI auto-contouring application (Seoul-based Oncosoft) that detects the treatment region and generates OAR contours on CT/MR, with its own brush-based editor and DICOM/HL7-FHIR interchange. Reported clinical deployment across multiple hospitals in South Korea.

When to choose it: sites wanting a fast standalone tool with an own editor and DICOM handoff to the TPS.

When to avoid it: if you require native TPS scripting or CBCT/PET contouring.

What it does. An excellent Brazilian cloud tool from Silva Ray that uses last-generation deep-learning AI to auto-contour organs-at-risk on planning CT for head & neck, thorax and pelvis — 47 OAR structures in total (19 head & neck, 15 thorax, 13 pelvis). It runs on remote GPUs (no local GPU needed), is fully DICOM-integrated, and its models are trained on hundreds of real clinical CT cases, with native Portuguese support and stated LGPD compliance.

When to choose it: a Brazilian service that wants a national, cloud-based CT OAR tool with native Portuguese support and LGPD-aligned data handling, and prefers cloud inference with no local GPU to install or maintain.

When to avoid it: if you need MR/CBCT/PET support, automated tumour/GTV segmentation, an in-product collaborative editor, or fully on-prem/air-gapped deployment.