[Paper] End-user validation of BRIGHT with custom-developed graphical user interface applied to cervical cancer brachytherapy

Source: arXiv - 2602.16321v1

Overview

The paper presents a custom graphical user interface (GUI) that lets clinicians interact with BRIGHT—a multi‑objective optimisation engine for cervical cancer brachytherapy. By turning a complex algorithm into an intuitive, visual workflow, the authors demonstrate that BRIGHT can produce treatment plans that are at least as good as, and often better than, the current manual practice.

Key Contributions

• Dedicated GUI for BRIGHT: Enables plan navigation, side‑by‑side comparison, dose‑distribution visualisation, and on‑the‑fly adjustments.
• End‑user validation: A multidisciplinary brachytherapy team evaluated the system on ten real patient cases, mimicking routine clinical workflow.
• Usability assessment: Achieved an System Usability Scale (SUS) score of 83.3, classified as “excellent”.
• Clinical performance: In 5/10 cases BRIGHT outperformed the standard plan, matched in the other 5, and was preferred by clinicians in 8/10 cases.
• Open‑source potential: The GUI and underlying BRIGHT engine are built on modular, reproducible code that can be extended to other cancer sites or optimisation problems.

Methodology

1. Algorithm backbone (BRIGHT) – Uses the GOMEA (Gene‑Optimisation‑Method‑for‑Evolutionary‑Algorithms) heuristic to generate a Pareto front of treatment plans, each balancing tumor coverage against sparing of healthy organs.
2. GUI design – Implemented in Python (Qt for the front‑end, VTK for 3‑D dose visualisation). Core features:
   • Plan carousel to scroll through Pareto solutions.
   • Pairwise comparison view with overlay of dose maps.
   • Interactive sliders to tweak weighting of objectives and instantly re‑run the optimiser for rapid “what‑if” analysis.
3. Clinical simulation – Ten previously treated cervical cancer patients were re‑planned using BRIGHT + GUI. The team performed blinded, one‑on‑one comparisons between the new and the original clinical plans.
4. Usability testing – After the session, participants completed the SUS questionnaire, providing a quantitative measure of the interface’s learnability and efficiency.

Results & Findings

These results indicate that the GUI not only makes the optimisation engine accessible but also translates into tangible clinical benefits.

Practical Implications

• Faster treatment planning – The visual navigation cuts down the iterative trial‑and‑error that radiotherapy teams currently endure, potentially freeing up valuable staff time.
• Better decision support – By exposing the full trade‑off surface, clinicians can make more informed choices, especially in complex cases where organ‑at‑risk constraints are tight.
• Extensible platform – The modular architecture means developers can plug in other optimisation algorithms, add new imaging modalities, or adapt the interface for other brachytherapy sites (e.g., prostate, breast).
• Data‑driven quality assurance – The system logs every plan variant, enabling retrospective analysis and machine‑learning‑based improvement loops.
• Potential for remote planning – Because the GUI runs on standard workstations and communicates with a backend server, multi‑site collaborations or tele‑medicine workflows become feasible.

Limitations & Future Work

• Sample size – Validation was performed on only ten patients; larger, multi‑center studies are needed to confirm generalisability.
• Learning curve – While SUS scores were high, some users required training to interpret Pareto fronts correctly.
• Integration with existing TPS – The current prototype operates alongside, rather than inside, commercial treatment planning systems, which may hinder seamless adoption.
• Future directions include:
  • Embedding the GUI directly into commercial TPS APIs.
  • Automating the selection of “optimal” plans using clinician‑defined utility functions.
  • Extending the framework to incorporate real‑time imaging feedback during needle placement.

Bottom line: By wrapping a sophisticated multi‑objective optimiser in a clean, interactive GUI, the authors have taken a big step toward making AI‑assisted brachytherapy both usable and clinically valuable—a development that developers and healthcare technologists should watch closely.

Authors

• Leah R. M. Dickhoff
• Ellen M. Kerkhof
• Heloisa H. Deuzeman
• Laura A. Velema
• Stephanie M. de Boer
• Lavinia A. L. Verhagen
• Danique L. J. Barten
• Bradley R. Pieters
• Lukas J. A. Stalpers
• Renzo J. Scholman
• Pedro M. Matos
• Anton Bouter
• Carien L. Creutzberg
• Peter A. N. Bosman
• Tanja Alderliesten

Paper Information

• arXiv ID: 2602.16321v1
• Categories: cs.NE
• Published: February 18, 2026
• PDF: Download PDF