2.D. Output measurements for validation

To test the models for accuracy and compare them against each other, the outputs of 272 sets of R and M covering the 24 options are measured. These are selected from treatment plans (for typical clinically used R and M combinations) as well as random sets of R and M (some of which are rarely selected for clinical beams, but are measured to cover a wide array of R and M combinations) and are measured independently of the model fit data points. Most of the points of measurement for the validation are located at the center of the SOBP and the isocenter. However, some measurement points are away from the center of the SOBP (maximum ± 5 cm offset of the middle of the SOBP) and/or the isocenter (60 measurements). The outputs predicted by the models are compared to the measured validation outputs (the percent difference = (Prediction – Measurement)/Measurement × 100%). Additionally, each model is compared against each other using the Student's t‐test.