Whole-volume diffusion kurtosis MR imaging histogram analysis of non-small cell lung cancer: Correlation with histopathology and degree of tumor differentiation

Ke Wang,Guangyao Wu
DOI: https://doi.org/10.1016/j.crad.2024.04.018
IF: 3.389
2024-01-01
Clinical Radiology
Abstract:AIM 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 streptavidin-perosidase method. Histogram analysis was performed using voxel-based raw data from each ROI. RESULTS NSCLCs were classified as grade 1, 2, and 3 according to differentiation degree. Histogram parameters of ADC and DKI could discriminate different grade of tumors (p<0.001). Receiver operating characteristic (ROC) curve analysis showed that Kapp 75th exhibited best performance with an AUC of 0.936 and sensitivity/specificity of 95.74%/80% (p<0.001) in distinguishing grade 1 from 2, ADC mean exhibited 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 ADC mean and 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.
What problem does this paper attempt to address?