Activity Number:
|
514
- Recent Advances in Imaging Statistics: Bayesian Methods and Beyond
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Imaging
|
Abstract #324316
|
View Presentation
|
Title:
|
A Study of Some Alternative Statistical Methods for the Characterization of Tumor Risk Based on PET Imaging Data
|
Author(s):
|
Eric Wolsztynski* and Tian Mou and Mark Muzi and Finbarr O'Sullivan
|
Companies:
|
University College Cork and University College Cork and UW-Radiology, Neurology & RadOnc and University College Cork
|
Keywords:
|
Cancer ;
Patient outcome ;
Positron Emission Tomography ;
Spatial modeling ;
Tumor heterogeneity ;
Texture
|
Abstract:
|
Diagnosis, therapeutic pathways and effectiveness are routinely assessed via Positron Emission Tomography (PET) in many prevalent and rare cancers. The size and nature of PET scans, usually under-exploited in quantitive manner, often have the ability to assess specific metabolic tumor features like avidity and heterogeneity. Two approaches are considered for use in this setting; spatial statistical models (including nonlinear regularized constructs) describing the volumetric distribution; and radiomics techniques extracted from marginal and joint characteristics of spatial distributions. Understanding interactions and relationships between these quantification alternatives may help with their use in multivariate prognostic models. The work explores the statistical validity and prognostic utility of alternative variables and their relationships for several cancers. It demonstrates the essential role of dedicated spatial modeling for useful metabolic characterization of tumors. Several case studies illustrate the difficulties and opportunities that combining techniques may bring to routine practice.
|
Authors who are presenting talks have a * after their name.