Program Information
Application of Data Analysis Techniques to Motion and Margin Assessment for Subgroup of Patients with Large Prostates
M Mamalui*, Z Li , University of Florida Health Proton Therapy Institute/Radiation Oncology, Jacksonville, FL
Presentations
SU-K-605-7 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 605
Purpose: To analyze intrafractional motion of patients treated for prostate cancer, aiming at detection of possible internal margin (IM) reduction for subgroup of patients with large/ultralarge prostates. Consideration for organs-at-risk adjacent to large/ultralarge prostates calls for possible margin reduction for this patient subgroup more frequently compared to patients at the opposite side of volume spectrum.
Methods: Patient positioning relies on 3-4 non-coplanar gold markers, localized on 2D orthogonal X-rays before and halftime during treatment. 78 records/2300datapoints were extracted from Record&VerifySystem(MosaiqTM), as well as corresponding anatomical parameters such as prostate volume (determined by MRI and TRUS) and BMI. In-house software analyzed daily intrafractional shifts. Using kmeans data clustering and Gaussian mixture model, dataset was partitioned into groups; clusters were examined for presence of correlations with prostate volume (two methods) and BMI. Implications for margin reduction were examined.
Results: Prostate volume from MRI was found best considered predictor of intrafractional shifts. GaussianMixtureModel application has consistently shown 3components with average 3D shift decreasing as prostate volume increased:(61cc,3mm);(93cc,1.9mm);(180cc,1.3mm). Means of shift data for small(<60cc)and large(>150 cc) prostate subgroups were 0.25 and 0.18cm. Entire dataset has not shown linear correlative behavior with considered predictive parameters. However,as dataset was clustered into subgroups,all-but-one data clusters have shown correlation (p<0.05)with prostate volume. This indicates that,for more detailed modeling,at least two predictors need to be considered,one of which is prostate volume by MRI. Interfractional means/standard deviations were calculated both for entire dataset and subgroup of prostate volumes >150cc. CTV-to-PTV margin from internal motion only was [vert=2.8mm,lat=2.3mm,long=2.6mm] based on entire dataset and reduced [vert=1.7mm,lat=1.8mm,long=2.0mm] for subgroup of large prostates. Largest reduction 1.1mm observed for AP anatomical direction.
Conclusion: Large dataset analysis techniques arehelpful in studying extensive treatment data collected in modern radiation therapy institutions. Analysis of prostate treatment intrafractional data suggest possibility for customized CTVtoPTV margin based on prostate volume.
Contact Email: