This study was approved by the Institutional Review Board and conducted in accordance with the ethical standards of the World Medical Association (Declaration of Helsinki). Thirty-one healthy adult subjects (20 males and 11 females; average age, 34.6 years; range, 25–52 years) were included in the study. None of the subjects had viral hepatitis and none were excessive drinkers. Written informed consent was obtained from all subjects before MRI. All subjects underwent US after MRI examinations. The MRI and US examinations were performed after the subjects fasted for at least 4 hours.

All examinations were conducted with the subject in the supine position using a 3T MRI system (Vantage, Toshiba Medical Systems, Otawara, Japan), equipped with a pair of phased array coils that had 16 channel outputs, placed at the front and back of the abdomen.

First, to locate the heart, stomach, spleen, liver, SMV, PV and SV, coronal 2D single-shot fast spin echo images were acquired (repetition time [TR]/echo time [TE] = 17910 ms/80 ms, flip angle [FA] = 90, field-of-view [FOV] = 400 × 350 mm², matrix size = 288 × 320, number of slices = 22–30, slice thickness = 4 mm, gap = 1 mm, acceleration factor = 2, acquisition time of 18 × 2 seconds during two breath-holds at the beginning of expiration) for Time-SLIP placement.

For non-contrast-enhanced MRA, 3D bSSFP with respiratory triggering was acquired in the coronal plane. Detailed tagged and untagged regions using Time-SLIP are shown in Figure 1. Respiratory triggering with an external belt sensing the respiratory cycle was conducted at the beginning of expiration using a bellows wrapped around the abdomen to reduce motion artifacts. All subjects were instructed to breathe calmly and regularly by listening to a recorded voice (“inhale … exhale”). The short tau inversion recovery method (TI = 220 ms) was used for fat suppression. A bSSFP scan was conducted with the following parameters: TR/TE/FA = 4.8 ms/2.4 ms/90° and receiver bandwidth = 781 Hz/pixel. TR and TE were the shortest, and an FA of 90° was the maximal value to clear the specific absorption rate limitation. The phase-encoding order was centric for both acquisition methods. Parameters related to spatial resolution were the same among sequences as follows: FOV = 330 × 200 mm², matrix size = 256 × 256, slice thickness = 2 mm, number of slices = 38–50 without a gap, number of acquisitions = 1. Parallel imaging was applied in the phase direction with a factor of two, resulting in the collection of 128 phase-encoding lines per respiratory-trigger in a centric order in both acquisitions. The actual spatial resolution was 1.3 × 0.8 × 2 mm³, and final images were reconstructed into an apparent spatial resolution of 0.6 × 0.4 × 1 mm³ with zero filling.

The LGV usually runs from the esophagogastric junction to the portal-SV junction or its vicinity. Two methods for LGV visualization with hepatopetal flow were as follows (Fig. 1). In the first method A, one Time-SLIP was placed on the whole abdomen to suppress signals using TI = 1513 ms. The other Time-SLIP was placed on the spatial area from the gastric fornix to the gastric upper body to recover signals of the area, which was considered the source of the LGV. The Time-SLIP thickness was 80 mm, and TI was 1500 ms. In the second method B, nonselective inversion recovery pulse inverts all magnetization in the region using TI = 1513 ms. One Time-SLIP was placed on the center of the esophagogastric junction in the axial plane to recover signals of the area, which was considered as the LGV source. The Time-SLIP thickness was 30 mm, and TI was 1500 ms.

When the LGV flow is hepatofugal, it is supplied by the SMV and IMV flow and/or SV flow. First, in the method for visualizing the LGV with hepatofugal flow supplied by SMV and IMV (method C), nonselective inversion recovery pulse inverts all magnetization in the region using TI = 1513 ms. One Time-SLIP was placed on the lower abdomen to recover the inflow signal from the mesenteric veins to PV. The Time-SLIP thickness was 150 mm, and TI was 1500 ms. Second, it is necessary to place a Time-SLIP at the spatial area containing the spleen to visualize the LGV with hepatofugal flow supplied from SV (12). However, when the set tagged region contained the spleen, it inevitably included part of the spatial area from the gastric fornix to the gastric upper body, which is a hepatopetal LGV flow source. Therefore, we tried to visualize veins that were distributed around the gastric fornix, cardia, and esophagus derived from the LGV with hepatofugal flow by placing the tagged region on the SV and LGV away from the gastric fornix to the gastric upper body (method D). As the method for visualizing those veins, nonselective inversion recovery pulse inverts all magnetization in the region using TI = 1513 ms. One Time-SLIP was placed on the spatial area containing the lesser curvature of the stomach, excluding the gastric fornix to the gastric upper body. TI was 1500 ms.

All measurements were performed using an ultrasound system (Aplio™ XG, Toshiba Medical Systems) by the same physician to avoid interobserver errors. A convex transducer with a center frequency of 3.75 MHz (PVT-375BT, Toshiba Medical Systems) extended operating frequency range was used. The LGV generally accompanies the left gastric artery (LGA). The vessel accompanying the LGA, which was connected to the PV or SV, was considered as the LGA. Color Doppler US was performed using the color map with a standardized Doppler protocol to detect blood flow direction. The Doppler flow velocity range was set in the maximum range of 8–15 cm/s to detect the LGV flow signal because the blood flow velocity is slow (15). The LGV flow direction was measured during short time breath-holding.

Images using maximum intensity projection (MIP) and source images were evaluated to define the LGV with hepatopetal flow (Fig. 2). We estimated whether the vessel visualized by non-contrast-enhanced MRA was the LGV using anatomical information of the bSSFP without the Time-SLIP. The “visualized LGV”, was defined as the LGV identified in the suppressed signal area on source images; and “clearly visualized LGV”, as the LGV identified in the suppressed signal area on both MIP and source images. Subsequently, we evaluated the vein with method C or method D. Method D was used for veins around the fornix and cardia. When the LGV was visualized with method A or B, and not visualized with method C or D, the visualized vessel was defined as the LGV with hepatopetal flow. Two radiologists evaluated the images, and reached a consensus. All assessments were performed using a free available DICOM viewer (YAKAMI Software, Kyoto, Japan).

To exclude a possibility that the vessel selectively visualized in method A or B was not even partly in accordance with true LGV, we estimated whether the correlation was in agreement with the running course, flow direction of the LGV between non-contrast-enhanced MRA and US, on the basis of US findings.

Left gastric veins on method A and B were visualized in 30 (97%) and 24 (77%) (Fig. 3), and clearly imaged in 21 (68%) and 17 (55%) of 31 subjects, respectively. For only one subject in whom the LGV was not visualized by method A, the LGV was also not visualized by method B.

There were no visualized LGVs and veins distributed around the fornix, cardia, and esophagus in methods C and D.

The LGV was detected in 23 (74%) of 31 subjects. The LGV flow direction was hepatopetal in 22 of 23 subjects. In one subject, The LGV could be detected, but the LGV flow direction could not be determined.

The LGV was visualized in all 22 subjects by method A, and the LGV with hepatopetal flow was detected by US. In all the subjects, the running course of LGV was similar in method A and US. In one subject, The LGV was not visualized by method A and US was unable to confirm the flow direction (Fig. 4).