Vasculogenic mimicry (VM), accounting for 17.2% of hepatocellular carcinoma patients¹, refers to tumor cells mimicking endothelial cells and directly participating in blood vessel formation², which is associated with tumor aggressiveness and poor clinical outcome of liver cancer¹. VM appears in 2 distinctive forms: the tubular type, with vascular lakes lined exclusively by highly invasive and genetically dysregulated tumor cells, and the patterned matrix type, with tumoral sinusoids swathed of extracellular matrix sheaths and surrounded by cancer cells which connect to endothelial cell lined blood vessels². Vascular disrupting agents (VDAs) can specifically target the abnormal cytoskeleton in tumor endothelial cells, rapidly disrupt the existing tumor vasculature in various solid tumors and, consequently, lead to central hemorrhagic necrosis in solid tumors^3,4. Dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI) is capable of non-invasive quantification of vessel permeability and perfusion^3,4; diffusion weighted imaging (DWI) allows detection of the Brownian motion of water molecules, which can be used as a parameter reflecting tissue viability and necrosis⁵. The aim of this study was two-fold: first, to investigate multiple MR imaging parameters for detection and characterization of different patterns of tumor vasculature in rat primary liver cancer model; and second, to explore the potential sensitive imaging biomarkers for early prediction of response to vascular disrupting therapy on primary live cancer.Study design: Multifocal primary liver cancer model was constructed in male Sprague-Dawley (SD) rats by diethylnitrosamine (DENA) gavage feeding. Tumor growth was monitored by T2-weighted and T1-weighted images (T2WI, T1WI) until the largest tumor lesion reached 8 mm. Rats were then received Combretastatin A4 phosphate (CA4P) intravenously at 10 mg/kg. DWI and DCE-MRI were performed both pre- (baseline) and post- (1 hr and 12 hrs) treatment. Image acquisition and analysis: MRI was performed on a clinical 3.0T MRI scanner (Trio, Siemens, Erlangen, Germany) and a human wrist coil (MRI Devices, Waukesha, WI). The following parameters were used: T2WI (turbo spin echo turbo spin echo sequence (TSE), TR/TE = 4000/70 ms, flip angle 150°, FOV 75 × 56.25 mm², matrix 512 × 384); T1WI (TSE, TR/TE = 626/15 ms, flip angle 160°, FOV 75 × 56.25 mm², matrix 512 × 384); DCE-MRI (three-dimensional (3D) T1-weighted sequence, TR/TE = 7/2.46 ms, flip angle 15°, FOV 61 × 88 mm², matrix 192 × 132) and a bolus of 0.02 mmol/kg Gd-DOTA (Gadoterate meglumine; Dotarem®, Guerbet, France) was intravenously injected after 30 pre-contrast measurements; DWI (2-dimensional SE echo-planar imaging (EPI) sequence, TR/TE = 3500/62 ms, flip angle 90°, FOV 73 × 135 mm², matrix 96 × 52) with 8 b values (0, 50, 100, 150, 400, 600, 800 and 1000 seconds/mm²). The apparent diffusion coefficient (ADC) was calculated by a monoexponential model. Histological analysis: Paraffin-embedded tumors samples were sliced into 5 μm followed by H&E staining.Figure 1 shows both VM types occurring in DENA-induced primary liver tumors. In the tubular type (Type I), non-endothelial cell-lined enlarged blood lake was identified and tumor cells appeared poor differentiated (angioma-like vascular lakes). In the patterned matrix type (Type II), sheaths of extracellular matrix form loops surrounding packets of cancer cells and tumoral sinusoids were lined by both endothelial cells and cancer cells inserted between these endothelial cells (mosaic vessels). In baseline, the tumor lesions of type I showed higher ACD_perf and ADC_diff compared with those in type II lesions. Figure 2 shows the different therapeutic responses to CA4P. The perfusion in type I lesions went up sequentially from 156 mm²/s at baseline to 219 and 264 mm²/s at 1 hr and 12 hrs post-treatment, while the tumor volume rapidly dropped by 11% at 1 hr post-treatment and partially recovered to 93% compare with the baseline at 12 hrs post-treatment, suggesting the transient and incomplete vascular shutdown effect upon CA4P treatment. In stark contrast to the type I lesions, the perfusion in type II lesions steadily dropped from 108 mm²/s at baseline to 105 and 94 mm²/s at 1 hr and 12 hrs post-treatment, indicating intratumoral necrosis was induced, while the tumor volume stayed nearly the same.DENA-induced primary liver cancer in rats appears an optimal model to investigate VM in hepatoma. DWI and DCE-MRI based multiparametric imaging biomarkers help to characterize and distinguish two different types of VM in primary liver cancers in rats. ACD_^perf, ADC_diff and T2WI-drived tumor volume sensitively predict the therapeutic responses of vascular disrupting agent CA4P on primary liver tumors.