Whole-volume diffusion kurtosis magnetic resonance (MR) imaging histogram analysis of non-small cell lung cancer: correlation with histopathology and degree of tumor differentiation.

ABSTRACT

AIMS: To evaluate the role of diffusion kurtosis imaging (DKI) histogram analysis in the characterization of non-small cell lung cancer (NSCLC) and to correlate DKI parameters with tumor cellularity. MATERIALS AND METHODS: Sixty-four patients with pathologically diagnosed NSCLCs were evaluated by DKI on a 3-T scanner. Regions of interest (ROIs) were drawn on the map of b1000 manually. All NSCLCs were histologically graded according to the degree of tumor differentiation. Tumor cellularity was measured by the nuclear-to-cytoplasm (N/C) ratio and the number of tumor cell nuclei (NTCN), the expression of Ki-67 was detected using the streptavidin-peroxidase method. Histogram analysis was performed using voxel-based on raw data from each ROI. RESULTS: NSCLCs were classified as grades 1, 2, and 3 according to differentiation degree. Histogram parameters of apparent diffusion coefficient (ADC) and DKI could discriminate between different grades of tumors (p<0.001). Receiver operating characteristic (ROC) curve analysis showed that Kapp 75th exhibited the best performance with an AUC of 0.936 and sensitivity/specificity of 95.74%/80% (p<0.001) in distinguishing grade 1 from grade 2, ADC mean exhibited the best performance with an AUC of 0.923 and sensitivity/specificity of 92.33%/86.67% (p<0.001) in distinguishing grade 2 from 3. N/C ratio and Ki-67 changed significantly with grade (p<0.01). Negative correlations were found between the ADC mean and the N/C ratio, Ki-67, Dapp mean and N/C ratio, whereas Kapp mean and N/C ratio, Ki-67 were positively correlated. CONCLUSIONS: DKI histogram analysis could quantitatively characterize NSCLC with different grades by probing non-Gaussian diffusion properties related to changes in the tumor microenvironment or tissue complexities in the tumor.