MRI and PET can provide tumor biology information noninvasively. Apparent diffusion coefficient (ADC) from diffusion weighted imaging (DWI) can represent cellularity, dynamic contrast enhanced (DCE) MRI can provide microcirculation, and fluorodeoxyglucose (FDG) PET can provide tumor metabolism. Intratumoral heterogeneity is often associated with adverse tumor biology^1,2 and it can be assessed by these imaging parameters. Thus, we investigated the relationships between intratumoral heterogeneity parameters derived from DWI, DCE MRI, and FDG PET in head and neck cancer.

Thirty patients who underwent DWI, DCE MRI, and FDG PET before treatment were retrospectively evaluated. We measured maximum and minimum ADC (ADCmax, ADCmin) values of primary tumor on DWI, Ktrans, Kep on DCE-MRI, and maximum, mean SUV, and peak SUL (lean body mass corrected SUV; SULp) on PET. For intratumoral heterogeneity evaluation, ADCdiff (ADCmax - ADCmin) was calculated on DWI MRI.^3,4 We calculated metabolism-perfusion parameters (SUVs/Ktrans, SUVs/Kep) using DCE MRI and FDG PET.⁵ First-order statistics (skewness, kurtosis, entropy) and high-order features (low grey-level run emphasis, high grey-level run emphasis, low grey-level zone emphasis) were calculated using texture analysis of PET.⁶ The relationships among heterogeneity parameters of primary tumor were assessed. ADCdiff, an index of heterogeneity on DWI, was significantly correlated with metabolism-perfusion parameters (SUVmean/Ktrans: r=0.381, p = 0.038; SULp/Ktrans: r= 0.368, p = 0.457; SULp/Kep: r = 0.407, p = 0.026) using DCE and PET. ADCdiff was significantly correlated with entropy (r = 0.458, p = 0.011) and high grey-level run emphasis (r = 0.428, p = 0.018) on PET texture analysis. We divided the patients to 2 groups by median value of ADCdiff. In the tumor showing high ADCdiff, SUVmean/Ktrans (p = 0.027), SUVmean/Kep (p = 0.015), SULp/Ktrans (p = 0.050), SULp/Kep (p = 0.035) were significantly increased compared with the tumor showing low ADCdiff. The rate of tumor recurrence or disease progression was significantly higher in high ADCdiff group than that of low ADCdiff group (66.7% vs. 26.7%, p = 0.028). Heterogeneous tumor on DWI showed metabolism-perfusion mismatch and poor prognosis in head and neck cancer. Tumor heterogeneity parameters derived from MRI and PET may provide valuable information on tumor biology and prediction of clinical outcome. A further study using a larger population of patients is needed to validate the clinical significance of intratumoral heterogeneity parameters derived from multimodality imaging.