Long-term direction
An original treatment-planning system, built one authorised capability at a time
VerityRT Studio is the multi-year programme: an independent, standards-based planning application covering the same categories of clinical work a modern planning system covers. It is a direction, not a product you can buy, and nothing on this page describes software that exists today.
Read this before anything else on the page
Two connected tracks
Waiting for a complete planning system before entering the market would be a mistake. The copilot funds and informs the platform.
- VerityRT Copilot
- Read-only integration, plan and chart QA, evidence, workflow, teaching and analytics. Reaches customers sooner with lower initial risk, and generates the workflow knowledge and validation data the larger programme needs.
- VerityRT Studio
- An original, standards-based, AI-native planning application. Can plan independently only after full commissioning and regulatory authorisation, capability by capability.
The dose engine is the hard part, and the honest answer is patience
Nobody should ask a language model to calculate dose, and nobody should imitate a proprietary algorithm. The sequence below is the one a physicist would recognise.
- 01Use dose imported from the existing planning system for the copilot and early Studio interface.
- 02Integrate a separately validated independent secondary-calculation engine for research and QA.
- 03Build or license a modular dose engine under qualified medical-physicist leadership.
- 04Start with one narrowly defined modality, energy range, machine class and technique.
- 05Commission against measured beam data, reference datasets, end-to-end phantoms and an accepted reference system.
- 06Lock the algorithm, material mapping, beam model, grid, calculation options and hardware environment.
- 07Validate heterogeneities, small fields, MLC behaviour, accessories, off-axis behaviour, buildup, oblique incidence and representative clinical geometries.
- 08Establish independent calculation and patient-specific QA requirements.
- 09Expand modality by modality, each with separate verification, clinical validation and regulatory review.
Optimisation, separated into layers
Every AI suggestion must be convertible into a visible, editable objective. A clinical tradeoff must never be hidden inside a score.
- Clinical intent
- Signed prescription, targets, organ priorities, protocol and documented exceptions.
- Objective compiler
- Converts approved intent into explicit mathematical objectives and constraints that a planner can read.
- Solver
- Deterministic, reproducible numerical optimisation.
- AI warm start
- Predicts useful initial objectives, fluence, geometry or an achievable tradeoff region. It starts the search; it does not end it.
- Deliverability
- Converts the solution into machine-deliverable parameters, with collision and clearance preview.
- Dose feedback
- Recalculates and updates the objective state.
- Independent evaluator
- Scores clinical goals, deliverability, robustness and anomalies using a separate implementation.
What the prototype actually shows today
- Imaging
- A procedurally generated synthetic volume behind a viewer adapter, so a validated viewer can replace it without touching a workspace.
- Contours
- Fixture structures with uncertainty and comparison metrics. No segmentation model runs.
- Beams and objectives
- Fixture geometry and objectives with stated reasons. No optimisation runs.
- Dose
- Deterministic DVH and metric fixtures. No dose calculation runs, and no number is generated at request time.
- Agents
- Typed output envelopes with status, uncertainty, evidence, versions and audit identity. The contracts are real; the outputs are fixtures.