Automated and Robust PERCIST-based Thresholding framework for whole body PET-CT studies

Abstract

Positron emission tomography (PET) is unique for quantitatively assessing treatment response before marked morphological changes are detectable by Computed Tomography (CT). PET response criterion (PERCIST) is a widely accepted approach of assessing metabolic response of malignant lesions by using Standardized uptake value (SUV) normalized by lean body mass (LBM) with a volume of interest (VOI) reference defined in the right lobe of liver. However, the operator-dependent delineation of VOI reference is a time consuming and subjective task. Although the VOI reference can be estimated from the co-aligned CT, the low-dose CT data in PET-CT poses challenge in liver segmentation. In this study, we propose a fully automatic framework to calculate the PERCIST-based thresholding for whole-body PET-CT studies. The framework consists of multi-atlas registration and voxel classification for CT data to segment liver structure and delineate the VOI reference, which is then mapped to the PET data to derive the value of SUVLBM thresholding for PET to select regions of high metabolism. We evaluated our framework with 28 clinical studies diagnosed with lung cancer or lymphoma, and demonstrated both reliability and efficiency in depicting lesions using PERCIST thresholding.