betaFeedback

WorksPhysics and Astronomy

Towards consensus on practical probabilistic evaluation metrics for clinical decision making in radiotherapy

Abstract

Introduction: The main aim of this paper is to identify clinically feasible probabilistic evaluation approaches for radiotherapy and to assess whether they can support clinical decision making better than current robustness evaluation approaches. To this end, we calibrate probabilistic clinical goals and identify evaluation metrics with a team of medical physics experts and apply these to three patient cohorts. Methods: Sixty intensity-modulated proton therapy plans from three Dutch proton centers were included: 20 neuro-oncological, 20 head-and-neck, and 20 breast cancer cases. Probabilistic evaluation was performed using polynomial chaos expansion with 10,000 simulated fractionated treatments per case. Volumetric and dosimetric CTV goals were calibrated at a 90% probability level. Selected probabilistic evaluation metrics were applied to representative patient cases and eight radiation oncologists participated in semi-structured interviews addressing preferred metrics and probability levels. Results: The calibrated volumetric goals were V0.95Dpres = 99.95%, 99.80% and 99.54% at a 90% probability level for the neuro-oncological, head-and-neck, and breast cohorts, respectively. The dosimetric goals were D98%,CTV = 0.973Dpres, 0.968Dpres and 0.967Dpres at a 90% probability level. All radiation oncologists considered the probabilistic evaluation metrics suitable for clinical decision making, and most regarded them as an improvement over the currently used VWmin/max approach. Preferences for individual metrics and probability levels varied among radiation oncologists, centers and treatment sites. Conclusion: Probabilistic evaluation metrics are expected to be clinically usable and an improvement over current methods by the interviewed radiation oncologists. Further consensus and clinical validation are needed before widespread clinical implementation.

Paper details

vlnc:W_dgzttd39
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…