MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Ensemble Learning Methods to Assess Dosimetric Predictors of Patient-Reported Toxicities After Prostate Stereotactic Body Radiotherapy

X Pan1,2*, J Huang1 , H Yu1 , X Qi3 , (1)School of Computer Science and Technology, Xi'an University of Posts and Telecommunications, Xi'an, Shaanxi 710121, PR China (2)Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing, Xi'an University of Posts and Telecommunications, Xi'an, Shaanxi 710121, PR China (3)Department of Radiation Oncology, University of California Los Angeles, CA 90095, USA

Presentations

(Wednesday, 7/17/2019) 9:30 AM - 10:00 AM

Room: Exhibit Hall | Forum 2

Purpose: To examine dosimetric correlations to patient-reported Quality-of-Life (QOL) for prostate stereotactic body radiotherapy (SBRT) using ensemble learning methods.

Methods: We analyzed 57 prostate patients who underwent SBRT treatment (40 Gy/5 fractions). Patient specific full spectrum DVH (in 10 cGy increment dose bin) for bowel and bladder were extracted from Eclipse planning system. The patient-reported QOL were assessed based on the Expanded Prostate Cancer Index Composite (EPIC-26) at the time between consultation and initial treatment (baseline) and every 3 months follow-up time. The ensemble ML
methods: Gradient boosted decision tree (GBDT), Adaboost and Random Forest (RF), were utilized to predict patient-reported QOL score changes in urinary and bowel domains respectively. These learning methods perform the classification task by building and combining multiple learners, achieving generalization performance that is significantly superior to a single learner. Support Vector machine (SVM), expecting to have excellent generalization capabilities based on small datasets, was also used for cross comparison. The plausible dosimeric metrics that are most associated with QOL score changes were further identified for bladder and rectal functions. Five-fold cross validation method was utilized to evaluate our findings.

Results: Among full-range of DVH points, the V32, V36, V40 and the volume receiving greater than 40.8 Gy were top ranking metrics that may be mostly associated with QOL score changes in urinary domain. At 1-year follow-up time, dosimetric quantities prognosticated patient reported urinary toxicities with the AUC of 0.90. While for rectal toxicities, there is no strong dosimetric correlations (AUC=0.60) were seen.

Conclusion: We demonstrate the use of ensemble learning to access full-range of dosimetric quantities and their associations with patient-reported outcomes. The set of DVH points may be highly correlated with patient-reported urinary toxicities. The identified DVH metrics can be considered as objective planning guidelines to further reduce patient reported toxicity for prostate SBRT.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by the National Natural Science Foundation of China (Grant No. 61702414), and the special funds for key disciplines in Shaanxi Universities and colleges.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email