Novel Constraint-Based Method to Evaluate and Score Patient Radiotherapy Plans
Abstract
Purpose
Conventional radiotherapy plan evaluation relies on isolated dose–volume metrics that provide limited insight into the magnitude and clinical significance of deviations from protocol-defined organ-at-risk (OAR) objectives. We propose a plan-level OAR dose deviation metric to quantify these deviations and enable systematic plan scoring.
Methods
We analyzed 288 left-breast and 294 right-breast approved radiotherapy plans; each prescribed 42.56 Gy in 16 fractions. For each OAR and objective, a constraint margin was defined as the difference between achieved and protocol-defined values, with positive values indicating increased margin. A relative constraint margin was obtained by normalizing to the objective target, enabling comparison across objectives. For each OAR and objective, relative margin values below 25th percentile were classified as most challenging, while values above 75th percentile were classified as least challenging. A plan-level score is computed by averaging relative constraint margins across all objectives and mapping the result to a percentile scale. This score is compared with the conventional plan score, defined as the fraction of objectives satisfied per plan. Paired differences between the two scores were analyzed using ECDF to assess whether the new metric systematically improves discrimination of plan quality.
Results
The breast PTV V40.43Gy objective was identified as the most challenging. In left-breast plans, the heart and left lung were consistently identified as the most challenging OARs, reflecting anatomical proximity and dosimetric trade-offs. ECDF analysis of paired scores showed that 60% of plans had higher scores under the constraint-aware metric, with a median difference of 1.04 (IQR: -1.85 to 3.68), indicating a systematic rightward shift relative to conventional scoring.
Conclusion
Constraint-aware plan-quality scoring quantifies deviations from protocol-defined targets and identifies the most and least challenging objectives and OARs. Normalizing constraint margins enables consistent, comparative evaluation of plan quality across patients and plan types, providing improved discrimination beyond conventional pass/fail approaches.