The implementation of RapidPlan in predicting deep inspiration breath-hold candidates with left-sided breast cancer.

The aim of this study is to determine if RapidPlan (RP) can be used as a prediction method to determine which left-sided supine breast cancer patients would benefit from the deep inspiration breath-hold (DIBH) technique. An RP model database was created with 72 clinically approved 3D conformal radiation therapy (3D-CRT) treatment plans. This model was validated by introducing 10 new patient data sets, creating RP-generated plans and comparing the clinically approved plan for the corresponding patient. The prediction ability of the model was then tested on the free-breathing (FB) scans of patients with clinically approved DIBH plans totaling 29 patients and results were then compared to the FB clinical plan attempts. A statistical analysis performed on the data indicated a strong correlation for the mean heart dose (R2 = 0.914; p-value < 0.001) with a standard deviation of 48.6 cGy. After validating the link between physician PTV and mean heart dose, the model was tested clinically on 15 patients by inserting "Test PTV Evals" that were contoured by the researchers as a surrogate for predicting mean heart dose. Statistical analysis showed a strong correlation between the dose to 5% of the heart (D5) and the mean heart dose (R2 values of 0.913 and 0.881, respectively) with a standard deviation for the mean heart dose of 27.2 cGy. It was concluded that by using a Test PTV Eval, the RP-generated plans were able to predict mean heart doses within ± 30.0 cGy.