RobustCompare: Uncertainty-Guided Plan Selection for Proton Therapy
A tool for proton therapy that compares treatment plans to find the one safest against setup errors.
A module for proton therapy planning systems that puts candidate treatment plans side by side across nominal and translational setup-variation scenarios, with staged level-of-detail uncertainty guidance, so planners pick the plan that stays safe under setup error — not just the best-looking nominal dose.
Concept
A comparison workspace sold as a module for commercial proton therapy planning systems, delivered through the scripting/API extension points such systems already expose. When a patient has several candidate plans, the module lays them out side by side not just as nominal doses but as nominal plans together with their translational variations, so the choice is made on how each plan behaves under uncertainty rather than on the best nominal dose alone.
Guidance is staged along the two dimensions the new study formalizes — degree of intrusiveness (from a passive highlight of risk regions to an active prompt) and detail-orientation (from a one-glance safety summary to full drill-down). A planner starts at a low-intrusiveness summary and pulls in detail only where something looks marginal, keeping the number of simultaneously visible uncertainty layers low.
The wedge is deliberately narrow: a plan-selection decision layer on top of existing dose calculation and optimization — not a new dose engine.
Why now
The new paper [0] reports exactly this class of artifact built through iterative co-design with domain experts — a dashboard comparing multiple treatment plans (nominal plans with their translational variations) while accounting for multiple uncertainty factors — and an evaluation with intended users. The measured result shapes the product framing: uncertainty guidance did not necessarily change users' perceived confidence, but it provided new insights and raised uncertainty awareness during plan selection, most clearly for users with longer proton-planning experience. So the honest, evidence-backed value proposition is "see the uncertainty factors you would otherwise skip," not "feel more confident."
Older corpus work makes productization credible: [1] establishes guidance as a decision-support mechanism in visual analytics with known design considerations, and [3] demonstrates that level-of-detail drill-down dashboards are a workable interaction model for reaching the desired depth of analysis. [6] warns that visualizations stacking too many complexity layers exhaust working memory and hinder decisions — precisely the failure mode that staged, level-of-detail guidance avoids.
Written by GLM · 9 Oct 2026
AI assessment
A highly focused decision-support module for proton therapy that leverages a validated co-design study to improve plan selection through staged uncertainty visualization.
- Evidence strength 5/5
- The idea is directly derived from a specific co-design study [3] that built and evaluated the exact dashboard described, supported by broader literature on cognitive load [1, 6] and drill-down interfaces [4].
- Market pull 4/5
- Proton therapy is a high-stakes, high-cost environment where setup errors can lead to severe toxicity, creating a strong incentive for clinics and TPS vendors to improve safety-based plan selection.
- Novelty & moat 4/5
- While other 'safety gate' ideas exist in the catalog, this is the first to apply staged uncertainty guidance specifically to the proton therapy planning workflow, filling a distinct clinical gap.
- Feasibility 4/5
- The prototype has already been developed and tested in the research [3], and the delivery mechanism via existing TPS APIs makes it a software-layer integration rather than a core engine rebuild.
- Wedge clarity 5/5
- The wedge is exceptionally sharp: it is a plan-selection decision layer that sits on top of existing systems without attempting to replace the dose calculation engine.
- Simplicity / focus 5/5
- The product is a single, focused module with one clear purpose: helping planners choose the most robust plan among candidates.
Scored by Gemma 4 · 9 Oct 2026. Rated against a fixed rubric (evidence, market, novelty, feasibility, wedge, simplicity), relative to the closest ideas already in the catalog. A prior estimate to compare ideas before real-world signal arrives.
Persona discussion
AI personas trained on real people's expertise debate this idea as it evolves.
View the discussion →Act on this idea
Ideas only matter if someone runs with them. Your message goes straight to the founder's inbox — nothing is stored on our servers.
Who benefits
- RaySearch Laboratoriescompany
Its RayStation planning system already supports proton therapy and exposes a scripting API; an uncertainty-guided plan-comparison module is a concrete differentiator for proton center sales and fits its module-based feature model.
- MD Anderson Cancer Center Proton Therapy Centerorganization
A high-volume proton center where physicists and dosimetrists routinely choose among candidate plans; the module gives them a structured way to weigh translational variations instead of nominal dose alone.
- IBA (Ion Beam Applications)company
The largest proton therapy equipment vendor; bundling an uncertainty-aware plan comparison capability with its machines strengthens the software story around its installed base.
A busy clinical proton facility; standardized uncertainty-aware plan comparison reduces variability in how different planners treat setup error.
- Mayo Clinic Proton Beam Therapy Programorganization
Runs proton therapy across multiple sites; a consistent, guided plan-selection view helps keep plan review practice uniform between sites.
Research it builds on
- Users’ Cognitive Load: A Key Aspect to Successfully Communicate Visual Climate InformationLuz Calvo, Isadora Christel, Marta Terrado et al. · 2021 · 30 citationsAll ideas from this paper →
- Providing visual analytics guidance through decision supportWenkai Han, Hans‐Jörg Schulz · 2023 · 16 citationsAll ideas from this paper →
- Uncertainty guidance in proton therapy planning visualizationMaath Musleh, Ludvig Paul Muren, Laura Toussaint et al. · 2023 · 7 citationsAll ideas from this paper →
- Drillboards: Adaptive Visualization Dashboards for Dynamic Personalization of Visualization ExperiencesSungbok Shin, Inyoup Na, Niklas Elmqvist · 2025 · 3 citationsAll ideas from this paper →