UCSF

Purpose

To evaluate the performance of diffusion-weighted imaging (DWI) and T2-weighted single shot fast spin-echo (SSFSE) imaging of the liver in the detection of hepatocellular carcinoma (HCC) in reference to the LI-RADS classification system.

Methods

MR images of 40 patients with 68 LI-RADS grade 3-5 lesions were analyzed. Two readers independently reviewed sequences and characterized lesion signal intensity, followed by consensus evaluation. CE-MRI served as reference standard. Sensitivities were compared across sequences. Lesion-to-liver contrast-to-noise ratios (CNRs) and apparent diffusion coefficients (ADCs) were measured and compared using the Wilcoxon signed-rank test across sequences and the Mann-Whitney U or Kruskal-Wallis test between LI-RADS categories. Inter-reader variability was assessed using Cohen's kappa statistic.

Results

Consensus sensitivities of LI-RADS 3-5 lesions using SSFSE images versus DWI were similar (0.53-0.63, p=0.089), however, the sensitivity with DWI b=700 was higher (0.63) than DWI b=0 (0.53, p=0.039). Lesion-to-liver CNRs were larger for all DWI sequences compared to SSFSE images (p<0.001 for all). ADCs of large (>2cm) LIRADS 3-5 lesions were lower than those of small lesions (1.09±0.33 vs. 1.31±0.26, p=0.02), however lesion ADCs were not different from those of adjacent hepatic parenchyma for any LI-RADS lesion.

Conclusions

DWI has a similar sensitivity compared to SSFSE, but intensity on DWI likely represents intrinsic T2 signal hyper-intensity rather than restricted diffusion as the ADC values were not lower than adjacent parenchyma. Therefore it may not be appropriate to consider hyper-intensity on high b-value as a separate ancillary criteria to T2 hyper-intensity in LI-RADS.