Nasopharyngeal carcinoma lymph nodes evaluation during therapy through diffusionweighted magnetic resonance images

Abstract

In order to assess the evolution of lesions during therapy, radiologists have to integrate information arising from different imaging techniques (e.g. morphological and functional) and time points (e.g. pre and post treatment). The purpose of this study was to quantitatively evaluate the treatment effect of chemotherapy on nasopharyngeal pathological carcinoma (NPC) using a semi-automatic framework integrating these data. The model concerns three main phases: 1) coregistration of volumes arising from the same time point on a common 3D uniform mesh; 2) pathological lymph nodes (PLN) identification in pre and post treatment volumes (PRE, POST); 3) tissues evaluation through apparent diffusion coefficient (ADC). The output of this procedure was compared to a hand made definition, performed on 12 patients longitudinally scanned. Each patient had a strong reduction of the NPC volume as reported by both qualitative and quantitative evaluations (Wilcoxon rank sum test P=0.885), the average volume decrease computed by the radiologist and by the semi-automatic segmentation procedure were: −79.21%±21.49% and −78.26%± 21.23% and the mean magnitude difference among them was 2.60% ± 2.02%. Although therapy induces morphological and functional changes on NPC lesions, this framework demonstrates a high sensibility and specificity.