A combination of MRI diffusion-derived vessel density (DDVD) and slow diffusion coefficient (SDC) can reliably diagnose liver hemangioma: a testing of three centers' data.

BACKGROUND

Magnetic resonance diffusion-derived vessel density (DDVD) is a physiological surrogate of the functional area of microvessels per unit tissue area, calculated according to: DDVD(b0b2) = Sb0/ROIarea0 - Sb2/ROIarea2, where Sb0 and Sb2 refer to the tissue signal when b is 0 or 2 (s/mm2). Sb2 and ROIarea2 can also be approximated by other low b-values (such as b=10 s/mm2) diffusion-weighted images. Slow diffusion coefficient (SDC) measures tissue slow diffusion, calculated according to: SDC = [S(b 1) - S(b 2)]/(b 2 - b 1), where b 1 and b 2 refer to a high b-value and a higher b-value, respectively, and S(b 1) and S(b 2) denote the diffusion-weighted image signal intensity acquired at the high b-value and the higher b-value. In this study, we studied whether a combination of DDVD pixelwise map (DDVDm) and SDC pixelwise map (SDCm) can reliably separate liver hemangiomas (HGs) from liver mass-forming lesions (MFLs).

METHODS

Three liver diffusion-weighted magnetic resonance imaging (MRI) datasets were tested. Dataset-1 consisted of 17 HGs, 35 hepatocellular carcinomas (HCCs), 6 intra-hepatic cholangiocarcinomas (ICCs), 2 metastases (Mets), and 7 focal nodular hyperplasias (FNHs). Images were acquired at 3.0-T, and DDVD was calculated with b=0 and b=2 s/mm2 images, SDC was calculated with b=400 and b=600 s/mm2 images. Dataset-2 consisted of 7 HGs, 56 HCCs, 4 ICCs, 14 Mets, and 8 FNHs. Images were acquired at 3.0-T, and DDVD was calculated with b=0 and b=10 s/mm2 images, SDC was calculated with b=500 and b=800 s/mm2 images. Dataset-3 consisted of 8 HGs, 12 HCCs, 2 ICCs, 13 Mets, and 1 FNH. Images were acquired at 1.5-T, and DDVD was calculated with b=0 and b=50 s/mm2 images, SDC was calculated with b=700 and b=900 s/mm2 images. Two readers jointly read all the images and made a decision regarding whether HGs could be separated from MFLs.

RESULTS

Liver HG typically showed high signal on the DDVDm and very high liquid signal on the SDCm. MFL typically showed iso- or slightly high signal on DDVDm and lower than liquid signal on SDCm. For the combination of dataset-1 and dataset-2, 95.8% (23/24) of the HGs and 97.7% (129/132) of the MFLs were correctly classified with confidence. One HG and two MFLs were correctly classified but without confidence, only one MFL was classified as undecided. Dataset-3 overall had a low signal-to-noise ratio, and 75.0% (6/8) of the HGs and 96.4% (27/28) of the MFLs were correctly suggested. One of the HGs was classified as undecided, while one HG was incorrectly suggested to be an MFL.

CONCLUSIONS

When a combination of DDVDm and SDCm is used to evaluate the liver, HG and MFL can mostly be reliably separated without the need for a contrast-enhanced scan. Our results will be relevant for HG confirmation when MRI is the first-line examination for the liver.