Patient Population
In this retrospective study,
we included patients with chronic liver disease who underwent surgical resection for focal liver lesion between June 2015 and June 2018 at Hepatobiliary and Transplant Surgery of University Hospital of Ancona.
All the lesions included were HCC at the surgical specimen.
All the patients had at least one triphasic contrast-enhanced CT or MRI within a month prior to surgery.
Patients with lack of clinical,
radiological or pathological data were excluded.
Imaging Protocol
Pre-surgical examinations were performed with a 64-slices TC (LightSpeed VCT,
GE Healthcare,
Milwaukee,
WI) or a 1.5 Tesla MRI (Signa HdXt,
GE Healthcare,
Milwaukee,
WI) with a triphasic post-contrastographic study.
MRI protocol included axial FSE T2w images with and without fat suppression,
a coronal T2w SSFSE sequence,
an axial T1w in- and out-of-phase GRE,
a DWI (b=0,
b=800 s/mm2).
The contrastographic study was obtained with administration of Gd-DOTA (0.1 mmol/Kg,
2 ml/s; Dotarem,
Guerbet, Roissy CdG Cedex, France) or Gd-EOB-DTPA (0.025 mmol/Kg,
1 ml/s; Primovist,
Bayer,
Berlin,
Germany).
3D spoiled,
fat-suppressed T1w GRE (LAVA) were used to acquire pre-contrast,
arterial,
portal venous,
late venous/transitional and eventual hepatobiliary phase.
CT examinations were performed with pre-contrast and post-contrastographic acquisitions of arterial (with bolus tracking),
portal venous and late venous phases with the following parameters: 120 kV,
modulated mA,
pitch 0.984:1,
iterative reconstruction (ASiR,
GE Healthcare,
Milwaukee,
WI),
reconstruction kernel Standard,
slice thickness/spacing 2.5/2.5 mm.
Image revision
Three readers with different experience rated in blind the presurgical imaging using LI-RADSv2018.
Reader 1 was a resident in radiology with 5 years of experience,
reader 2 had 10 years of experience and reader 3 had 30 years of experience.
LI-RADS Major and Ancillary Features were assigned for each lesion as in LI-RADSv2018 [13].
Statistical Analysis
Statistical analysis was performed with MedCalc v12.5 (MedCalc Software,
Ostend,
Belgium).
IRA was calculated using the Intraclass Correlation Coefficient (ICC).
For pathological correlation,
discordant LI-RADS Major or Ancillary Features were reviewed in consensus,
and Chi-square Test was used. Significant p was set at p<0.05.