Modelling DW-MRI data from primary and metastatic ovarian tumours.

Abstract

ObjectivesTo assess goodness-of-fit and repeatability of mono-exponential, stretched exponential and bi-exponential models of diffusion-weighted MRI (DW-MRI) data in primary and metastatic ovarian cancer.MethodsThirty-nine primary and metastatic lesions from thirty-one patients with stage III or IV ovarian cancer were examined before and after chemotherapy using DW-MRI with ten diffusion-weightings. The data were fitted with (a) a mono-exponential model to give the apparent diffusion coefficient (ADC), (b) a stretched exponential model to give the distributed diffusion coefficient (DDC) and stretching parameter (α), and (c) a bi-exponential model to give the diffusion coefficient (D), perfusion fraction (f) and pseudodiffusion coefficient (D*).ResultsCoefficients of variation, established from repeated baseline measurements, were: ADC 3.1 %, DDC 4.3 %, α 7.0 %, D 13.2 %, f 44.0 %, D* 165.1 %. The bi-exponential model was unsuitable in these data owing to poor repeatability. After excluding the bi-exponential model, analysis using Akaike Information Criteria showed that the stretched exponential model provided the better fit to the majority of pixels in 64 % of lesions.ConclusionsThe stretched exponential model provides the optimal fit to DW-MRI data from ovarian, omental and peritoneal lesions and lymph nodes in pre-treatment and post-treatment measurements with good repeatability.Key points• DW-MRI data in ovarian cancer show deviation from mono-exponential behaviour • Parameters derived from the stretched exponential model showed good repeatability (CV 7 %) • The bi-exponential model was unsuitable because of poor parameter repeatability • The stretched exponential model showed comparable repeatability to the mono-exponential model • The extra parameter (α) provides scope for investigation of heterogeneity or response