Association between radiomics of diffusion-weighted imaging and histopathology in hepatocellular carcinoma. A preliminary investigation

Objective

Diffusion-weighted imaging and the quantified apparent diffusion coefficient (ADC) correlate with cell density and histopathological features in tumors. Radiomics analysis may provide more insight into the underlying microstructure and may better correlate with histopathology. The present study used cross-sectional guided biopsy specimens to exploit the precise spatial localization of the performed biopsy to correlate radiomics features of the ADC map with immunohistochemical features in hepatocellular carcinoma (HCC).

Materials and methods

A total of 51 patients (11 female patients, 21.6 %) were included in the present study. The mean age was 71.9 ± 9.9 years, ranging from 42 to 91 years. Prebioptic liver MRI with diffusion-weighted imaging was used to correlate the radiomics features of the ADC maps with the immunohistochemical features quantified in liver biopsy. Proliferation potential Ki 67, leukocyte count and tumor-stroma ratio were evaluated as histopathological parameters.

Results

The following ADC texture features were correlated with the Ki 67 index _MinNorm (r = −0.307, p = 0.03), Vertl_RLNonUni (r = − 0.309, p = 0.03), 135dr_RLNonUni (r = −0.346, p = 0.01). The texture feature _MinNorm achieved the best diagnostic accuracy with an area under the curve of 0.76 (95 % CI 0.60–0.91, p < 0.01) to discriminate between low and high proliferative HCC. Multiple statistically significant correlations were found between ADC texture features and tumor-stroma-ratio, the highest for S(0,1)Contrast (r = 0.460, p = 0.001). No statistically significant correlations were found between the ADC texture features with the CD45+ leukocyte count and grading.

Conclusion

Radiomics features of the ADC maps can reflect the underlying histopathology in HCC patients including the proliferation potential and tumor-stroma ratio but not CD45 positive cells and tumor grading. The complex interactions between quantitative imaging and histopathology need to be further investigated in a validation cohort.