Unencrypted login | home

Program Information

Predicting Complication Probabilities Beyond the Dose-Volume Histogram: Dose-Mapping to Identify Critical Dose-Volume Risk Factors


J Deasy

J Deasy*, Z Saleh, A Apte, Memorial Sloan Kettering Cancer Center, NEW YORK, NY

TU-C-BRA-1 Tuesday 10:30:00 AM - 12:30:00 PM Room: Ballroom A

The goal of radiotherapy is to eradicate all tumor cells with regenerative capacity while keeping treatment-related complications to a manageable and subjectively tolerable acceptable level for each patient. Therefore it is paramount that we understand the factors that cause treatment related severe complications. Although clinical, genetic, and other biological factors play a role, dose-volume patterns are known to be the key driver of toxicity. Until recently, attempts to predict Normal Tissue Complication Probability (NTCP) for a given endpoint have relied almost exclusively on analyses that relate dose-volume-histogram (DVH) characteristics to the risk of complications through models such as the generalized equivalent uniform dose/Lyman-Kutcher-Burman model (Marks et al., (2010) 76:S10-S19.) However, it is known that function and repair processes are not uniformly distributed as independent processes throughout organs or tissues. Therefore, these DVH-based models discard spatial information that is likely to be important with respect to the development of toxicity. This talk will discuss methods and recent results (published and unpublished) that go beyond the DVH to predict toxicity in multiple ways, for example: by looking at imaging-based changes during a course of radiotherapy (), or by mapping dose-distributions from multiple patients to a reference patient to probe point-by-point correlations. These methods present exciting new avenues to potential identify more effective risk factors and resulting predictive models. However, they have their limits and require technical advances that will also be discussed, especially in the area of deformable image registration QA and statistical analysis.

Learning objectives include:
1. Understanding the process of mapping dose distributions to a reference anatomy to correlate with an outcome variable.
2. Understanding key uncertainties including deformable image registration, and a suggested method to evaluate this uncertainty.
3. Understanding the statistical limits of this method, including confounding variability.

This research was partially supported by NIH grant R01 CA85181.



Contact Email