Encrypted login | home

Program Information

Statistical DVH Dashboard: A Scalable Solution for Bringing Big Data to Treatment Plan Evaluation

C Mayo

C Mayo*, J Yao , D Litzenberg , M Kessler , M Matuszak , J Moran , J Balter , A Eisbruch , G Weyburn , R Ten Haken , Univ Michigan Medical Center, Ann Arbor, MI


TU-H-FS4-8 (Tuesday, August 1, 2017) 4:30 PM - 6:00 PM Room: Four Seasons 4

Purpose: Traditional plan evaluation methods do not include quantitative evaluation of dose volume histogram (DVH) curves or metrics with respect to historical plans. We developed statistical DVH-based metrics and visualization methods using Big Data to quantifying comparison of treatment plans to historical experience and among different institutions.

Methods: Per plan DVH curves plotted with statistical distribution summarizations of volume-normalized DVH curve sets were visualized in statistical DVH plots. Detailed distribution parameters were calculated and stored in JSON files to facilitate multi-institutional comparisons. In plan evaluation, structure DVH curves were scored against a computed statistical DVH and a Weighted Experience Score (WES). Individual clinically-used DVH-based metrics were integrated into Generalized Evaluation Metric (GEM), as a priority-weighted sum of normalized incomplete gamma functions. Historical treatment plans for 351 head and neck patients, 104 prostate patients treated with conventional fractionation, and 94 SBRT liver patients were analyzed to demonstrate the usage of statistical DVH, WES, and GEM in plan evaluation. A shareable dashboard plugin was created to display statistical DVH and integrate GEM and WES scores into clinical plan evaluation. Benchmarking with NTCP scores was carried out inter-comparing behavior of GEM and WES scores.

Results: DVH curves from historical treatment plans were characterized and presented, with difficult-to-spare structures (i.e. frequently compromised organ at risk OAR) identified. Quantitative evaluations by GEM and/or WES compared favorably with NTCP LKB model, transforming a set of discrete threshold-priority limits into a continuous model reflecting physician objectives and historical experience

Conclusion: Statistical DVH offered an easy-to-read detailed but comprehensive way to visualize quantitative comparison to historical experience and among multi-institutions. WES and GEM metrics offer flexible means to incorporate discrete threshold-prioritizations and historic context into a set of standardized scoring metrics. Together, they provide a practical approach for incorporating Big Data into clinical practice for treatment plan evaluation.

Contact Email: