Room: Karl Dean Ballroom C
Purpose: Online adaptive replanning (OLAR), though vital to account for interfractional variations during radiation therapy (RT), is time-consuming and labor-intensive when compared to the repositioning method. Repositioning is enough for minimal interfractional deformations. Therefore, determining indications for OLAR is desirable. We introduce a method to rapidly determine need for OLAR by analyzing the Jacobian determinant histogram (JDH) obtained from deformable image registration (DIR) between reference (planning) and daily images.
Methods: The proposed method is demonstrated on daily CTs acquired using an in-room CT during CT-guided RT for prostate cancer. DIR between daily and reference CTs was performed. JDHs were extracted from the prostate and a uniform 10mm expansion around it. Using machine learning, a classification tree was trained to determine the JDH metrics that can predict the need for OLAR for a daily CT set. Sixty daily CTs from 12 randomly selected prostate cases were used for the training dataset, with OLAR and repositioning based on fractional dosimetric plans as the classes. The resulting classification tree was tested using an independent dataset of 50 daily CTs from ten other patients with 5 CTs each. Accumulated doses from 5 fraction plans of different strategies were compared.
Results: Out of a total of 26 JDH metrics tested, 4 were identified as predictive as to whether OLAR was substantially superior to repositioning for a given fraction. The obtained metrics correctly identified all cases where benefits of OLAR were obvious. Borderline cases with marginal dosimetric benefits from OLAR were inconclusive. Accumulated doses with hybrid OLAR and repositioning (using classification tree) were comparable to all OLAR fractions.
Conclusion: Effective JDH metrics that can quickly determine the necessity of online replanning based on image of the day without segmentation have been identified using a machine learning process. These metrics can be implemented for online adaptive replanning.