Purpose Dynamic contrast-enhanced MRI (DCE MRI) had been proved valuable to ultra-early predict neoadjuvant chemotherapy (NAC) response in nasopharyngeal carcinoma (NPC) ¹. Another study suggested that diffusion parameters derived from Non-Gaussian model might reveal additional tissue properties of NPC as compared with mono-exponential model ² and we recently found that diffusion kurtosis imaging (DKI) was valuable for predicting NAC response in NPC ³. The purpose of this study was to determine the value of combining DCE MRI and DKI in early monitoring NAC response in NPC.

Materials and Methods Fifty-three newly diagnosed locally advanced NPC patients received four MRI exams using a 3.0 T Philips MRI system (Achieva TX, Best, the Netherlands) before and on the 4th, 20th and 40th days after NAC initiation prospectively. DKI with 4 b-values (0, 500, 1000, and 1500 mm2/sec) were applied. The DCE MRI parameters derived from extended Tofts’ model (including K_^trans [the volume transfer constant of Gd-DTPA], Κ_ep [flux rate constant], ν_e [the extracellular volume fraction of the imaged tissue], and ν_p [the blood volume fraction]) and DKI parameters (including D [corrected diffusion coefficient] and K [excess diffusion kurtosis coefficient]) were calculated from Philips software developed in IDL 6.3 (ITT Visual Information Solutions, Boulder, CO). Parameters and their corresponding changes Δparameter_(0-X) (X = 4, 20 or 40 days) were compared between responders and non-responders after one or two NAC cycles using independent student T test or Mann-Whitney U test followed by logistic regression and receiver operating characteristic curve (ROC) analyses.

Results Among 53 enrolled NPC patients, 15 and 24 subjects were categorized into responders group (RG) after one and two NAC cycles respectively. Compared to non-responders group (NRG), RG patients after either one or two NAC cycles presented significantly higher mean K^trans and D values at baseline and larger ΔK^trans_(0-4), Δv_p(0-4) and ΔD_(0-4) values (all P values <0.05). Moreover, K_ep and ΔK_ep(0-4) proved significantly larger (P <0.05) in RG than in NRG grouped by response after two NAC cycles but did not after one NAC cycle. For the above parameters, ROC analyses demonstrated that combined DCE-MRI and DKI model gained higher diagnosis accuracy to distinguish non-responders from responders than using either alone both after one (0.857 vs 0.749 or 0.812, Fig. 1) and two (0.987 vs 0.872 or 0.897, Fig. 2) NAC cycles.

Discussions The DKI and DCE MRI derived parameters could reflect physiological features and pathologic changes (diffusion and angiogenesis) at the micron level^1,4 and corresponding changes after therapy^1,3. In this pilot study, a combination of DKI and DCE MRI was explored and proved that a higher diagnosis accuracy was obtained in early predicting NAC response in NPC. The value of multi-parameters MRI in grading glioma had been validated recently⁵. This study demonstrate the application of multi-parameters MRI for NAC assessment in NPC was encouraging.

Conclusions Both DCE MRI and DKI demonstrated a trend toward higher diagnostic accuracy to early assess two cycles than one cycle NAC treatment in NPC. Combining DCE-MRI and DKI modalities gained higher accuracy for both assessing one and two NAC treatment than either model was used independently.