The ML-Based Patient-Specific Quality Assurance tool is being developed to support risk-informed assessment of IMRT, VMAT, SRS and SBRT treatment plans. The platform will analyse selected treatment-plan characteristics, modulation and complexity metrics, dosimetric information and available treatment-delivery or machine-log data.
Machine-learning models will be used to estimate relevant PSQA outcomes, such as gamma passing rate, plan isocentre dose agreement and the probability of a plan requiring additional investigation. Results will be presented with model-performance information, uncertainty indicators and the contributing plan features to support transparent clinical interpretation.
The tool is intended to complement established patient-specific QA procedures by helping medical physicists prioritize measurements, identify potentially unusual plans and support research into risk-adapted QA workflows. It will not independently approve treatment plans or replace measurement-based QA, secondary dose verification, professional review or institution-specific clinical procedures.
Clinical implementation will require formal commissioning, local validation, ongoing performance monitoring and approval under the institution’s quality-management programme.
What we can provide
- Support for IMRT, VMAT, SRS and SBRT plans
- Treatment-plan and plan-complexity feature analysis
- Optional treatment-delivery and machine-log features
- Gamma passing-rate prediction
- Plan isocentre dose-agreement prediction
- Plan-level QA risk classification
- Identification of unusual or out-of-distribution plans
- Model confidence and uncertainty indicators
- Feature-contribution and interpretability information
- Configurable institutional QA thresholds
- Model-performance monitoring
- Exportable and printable PSQA assessment summary
- Research-data export for approved studies