Evaluation of Six Diffusion-weighted MRI Models for Assessing Effects of Neoadjuvant Chemoradiation in Pancreatic Cancer Patients

Purpose: To compare 6 diffusion-weighted imaging (DWI) MRI models for response evaluation in patients with pancreatic ductal adenocarcinoma (PDAC). Materials and Methods: DWI images were acquired at 3T for b = 0-600 s/mm(2) in fourteen patients with advanced PDAC during 2 separate pretreatment sessions and 9 patients with (borderline) resectable PDAC pre and post neoadjuvant chemoradiation. Data was fitted with a mono-exponential (ADC), double mono-exponential to b = 0 and 100 s/mm(2) (ADC(fast)), and b = 100 and 600 s/mm(2) (ADC(slow)), IVIM model with D* free (D, f, D*) and fixed (D, f), tri-exponent (D, f(1), f(2)), and stretched exponent model (DDC, alpha). Goodness of fit (adjusted R-2), tumor to normal tissue contrast, repeatability (coefficient of variation), and parameter correlations (Spearman's rho) were assessed for the repeated measures. Treatment induced changes were assessed and compared to the repeatability. Results: The mono-exponential model had the lowest goodness of fit in both tumor (R-2 = 0.94) and normal-appearing pancreas (R-2 = 0.88). Tumour to normal tissue contrast was higher for the 'non-diffusion' parameters (ADC(fast), f, D*, f(1), f(2), alpha), with better repeatability for the diffusion parameters (ADC, ADC(slow), D, DDC). Diffusion parameters were strongly correlated between the models (rho >= 0.81) and showed a general treatment associated increase. All models were able to identify individual treatment effects, showing a change greater than the repeatability in 5 out of 9 patients for at least one of the parameters. Conclusions: Individual treatment evaluation is possible with all investigated DWI models, with treatment associated changes exceeding the repeatability. The double monoexponential fit with ADC(fast) and ADC(slow) is able to discriminate between non-diffusion and diffusion related effects, is measured fast and can be performed on most commercial scanners, making it an attractive alternative for the more advanced multi-parametric models in radiotherapy treatment evaluation. (C) 2018 Elsevier Inc. All rights reserved.