ESTRO 2024 - Abstract Book

S5129

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

2586

Digital Poster

Quantitative DCE-MRI in head and neck cancer: a comparison of different modelling approaches

Marte Kåstad Høiskar 1 , Oddbjørn Sæther 2 , Mirjam D Alsaker 3 , Kathrine R Redalen 1 , René M Winter 1

1 Norwegian University of Science and Technology, Department of Physics, Trondheim, Norway. 2 St. Olavs hospital, Trondheim University Hospital, Department of Radiology and Nuclear Medicine, Trondheim, Norway. 3 St. Olavs hospital, Trondheim University Hospital, Cancer Clinic, Trondheim, Norway

Purpose/Objective:

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a non-invasive method to probe tissue microvasculature and has demonstrated prognostic potential for radio-chemotherapy in head and neck cancer (HNC). However, a lack of consensus and standardization in quantitative DCE-MRI is hindering clinical use. Different pharmacokinetic models have shown promise for quantitative DCE-MRI analysis. Some of these models e.g., the Tofts model (TM), require an arterial input function (AIF) that describes the arterial concentration time course of the contrast agent post injection. The AIF is important for reliable estimation of the model parameters and is ideally measured for each patient individually, but when impeded, often a population AIF derived from the image data of other patients is used. While some studies use different implementations of the same model, such as different AIF methods, other studies use different models, such as the extended Tofts model (ETM) or Brix model (BM). This poses a challenge to meta-analysis and results in many potential candidates for biomarkers. The present study investigated accuracy and robustness of the population AIF, correlations between pharmacokinetic parameters across different models and their association to T stage and human papillomavirus (HPV) status for HNC. DCE-MRI was acquired in 44 patients with HNC before the start of radio- or radiochemotherapy. The study was approved by the regional ethics committee and all patients gave written informed consent. Arterial regions were delineated on the imaging data and used to determine individual AIFs. In addition, population AIFs were calculated with six different approaches. DCE-MRI data was analyzed in n=36 primary and n=44 lymph node tumors using TM with population and individual AIFs, ETM with individual AIFs, BM, area under the curve (AUC) for different times, and time to half peak (TTHP). Intraclass correlation (ICC) was used to examine the robustness of the population AIF by comparing the TM parameter values obtained with different population AIF. The concordance correlation coefficient (CCC) of TM parameter values was used to examine the accuracy of the population AIF compared to individual AIFs which served as reference. Pearson correlation coefficient (PCC) and Mann-Whitney U test were used, respectively, to explore correlations between all different model parameters and their association to T stage and HPV status. Material/Methods:

Results:

Population AIFs obtained with different methods are plotted in Figure 1. The population AIF was robust against method variations with an ICC of TM parameters ranging from 0.83 (K trans ) to 0.99 (v e ). However, parameter values differed when using individual AIFs as shown by a CCC of 0.49 to 0.55. PCCs for different parameter combinations are