Using ESWAN To Differentiate Hepatocellular Carcinomas From Cavernous Hemangiomas: A Preliminary Research

PurposeTo evaluate the feasibility of ESWAN post-processed multiple parameters todifferentiate hepatocellular carcinomas (HCC) from cavernous hemangiomas (CHL).Meterials and Methods164patients (mean age=54.85years±12.22(SD);104M:60F) admitted in our hospitalunderwent a ESWAN MR imaging on a1.5T HD propeller MR scanner (GE) with an8-channel phased array coil, including87patients with97pathologically proven HCClesions and77patients with103clinically or pathologically proven CHL lesions. Allpatients were examined with T1WI、T2WI、LAVA and ESWAN (axial3D, matrix:256×192, first echo: TR/TE:16.3/2.1ms,5echos, flip angle=15°). The phase,magnitude and R2*maps was generated from original ESWAN data. The signalintensity heterogenity in phase, magnitude and R2*maps of two groups was tested withChi-square test. And signal intensity heterogenity in phase, magnitude and R2*maps ofHCC was tested with Chi-square test. The values of phase, magnitude and R2*fromHCC and CHL groups were compared with the Mann-Whitney non-parametric U test. Acut-off value was evaluated with receiver operator characteristic (ROC) curve andsensitivity and specificity were calculated. The visible vessels score on phase,magnitude and R2*maps from HCC and CHL groups were compared with theMann-Whitney non-parametric U test according to visibility on LAVA images. Thedifferentiating ability of phase, magnitude and R2*maps on inter-tumoral andperi-tumoral vessels were compared in HCC and CHL groups with ROC curve. Observethe peri-tumoral ring on phase, magnitude and R2*maps, and compare the differencebetween HCC and CHL groups with Mann-Whitney non-parametric U test. Thecorrelation with ring on LAVA images and lesions size was calculated with Spearsoncorrelation analysis. The differentiating ability of LAVA, phase, magnitude and R2*maps on peri-tumoral rings were compared in HCC and CHL groups with ROC curve. The difference of inter-tumoral vessels, peri-tumoral vessels and rings on phase,magnitude and R2*maps between HCC and CHL groups were compared withnon-parametric Kruskal-Wallis H test.Results The signal was more heterogenous in HCC than that in CHL, especially onphase map. There was significant difference of phase, magnitude and R2*valuesbetween HCC and CHL group. And R2*had highest diagnosis value. A threshold of9.2978Hz for the minimum R2*value in the diagnosis of HCC had a sensitivity of94.8%, specificity of85.4%, positive predictive value (PPV)86.0%and negativepredictive value (NPV)94.6%. The inter-tumoral visible vessels score on phase andmagnitude maps of two groups was higher than that on R2*map, and there was nodifference. The area under curve (AUC) for diagnosing HCC was0.879and0.942onphase and magnitude maps, respectively. If diagnose HCC when score higher than3,the phase map had a sensitivity of90.7%, specificity of77.7%, PPV79.3%and NPV89.9%, and the magnitude map had a sensitivity of85.6%, specificity of92.2%, PPV91.2%and NPV87.2%.④The inter-tumoral vessel scores on magnitude and phasemaps was higher than that on R2*map in two groups and there was no difference. TheAUC for diagnosing CHL was0.517、0.666and0.707on phase, magnitude and R2*maps, respectively. If diagnose CHL when score lower than3, the magnitude map hadbetter diagnosis value with a sensitivity of75.7%, specificity of49.5%, PPV61.4%andNPV65.8%.⑤The peri-tumoral ring scores on phase map was higher than that onmagnitude and R2*maps in two groups, and there was no correlation between ringscores and lesions size, neither the capsule on LAVA image.⑥The AUC fordiagnosing CHL with peri-tumoral ring was0.270、0.601and0.766on phase,magnitude and R2*maps, respectively. If diagnose CHL when score lower than3, thephase map had better diagnosis value with a sensitivity of71.8%, specificity of75.3%,PPV75.5%and NPV71.6%.⑦There was significant difference of inter-tumoral,peri-tumoral vessels and peri-tumoral ring scores on ESWAN induced phase, magnitudeand R2*maps in two groups.Conclusion The phase and R2*value induced from ESWAN could be quantitativeparameters to differentiate HCC from CHL, and R2*value was more sensitive andreliable. The inter-and peri-tumoral vessels and peri-tumoral ring could be shown withESWAN sequence that provided more morphological information to differentiate HCCfrom CHL which conventional MR sequences could not do. ESWAN would be a more effective method to differentiate HCC from CHL without contrast injection for patientswith severe renal failure and contrast allergy.