Pancreatic perfusion and its response to glucose as measured by simultaneous PET/MRI

The study was approved by the Ethical Committee for Animal Research of the Uppsala Region and was performed according to the Uppsala University guidelines on animal experimentation (UFV 2007/724). High-health herd-certified normoglycemic male pigs (Yorkshire × Swedish Landrace x Hampshire, male, n = 4, weight 28.5–35 kg) were used. Anesthesia was induced by an IM injection of tiletamine–zolazepam (Zoletil Forte^® vet. 250 mg/mL) 5 mg/kg and medetomidine (Domitor^®vet. 1 mg/mL) 0.025 mg/kg and maintained by an IV infusion of ketamine (Ketamin^® 10 mg/mL) 28 mg/kg × h, midazolam (Midazolam Actavis^® 5 mg/mL) 0.1 mg/kg × h, and fentanyl (Fentanyl B. Braun^® 50 µg/mL) 3.5 µg/kg × h. A detailed description of the MRI compatible pig anesthesia procedure was previously published [12]. Venous catheters were placed in the left and right auricular vein for anesthesia infusion (left ear) or [¹⁵O]H₂O and glucose infusion (right ear). In one pig, an arterial Seldinger catheter was placed in the arteria carotis with the tip placed in the left cardiac ventricle for administration of microspheres. Additionally, an arterial catheter was also placed in the arteria femoralis to obtain a reference sample.

Pigs were examined up to three times during 1 day using a PET/MRI scanner (Signa, GE Healthcare). The instrument is equipped with a 3 T magnet and allows for simultaneous PET/MRI measurements.

Pancreas perfusion was assessed simultaneously and independently by [¹⁵O]H₂O PET and DWI, either at basal conditions or after IV administration of glucose. In one pig, perfusion was also assessed by injection of microspheres. Table 1 demonstrates the details on the experimental procedures.

The first PET/MRI examination aimed to measure the pancreatic perfusion at basal conditions. Before subsequent PET/MRI examinations, the pigs were given an IV glucose challenge (0.1 g/kg or 0.5 g/kg). Glucose was administered as a 2-min infusion of 10 or 50 mL solution (300 mg/mL), immediately before the PET/MRI examination. Blood glucose was sampled before and after each PET/MRI examination.

One pig was administered colored microspheres to measure the pancreatic and islet perfusion [2]. In this pig, 12.1 mL of a microsphere solution (approximately 10⁷ red microspheres, 15.5 ± 0.3 µm, EZ Trac, Los Angeles, CA, USA) was administered intra-arterially in the left cardiac ventricle 1 min after the start of the PET/MRI examination during basal conditions. An arterial reference sample was withdrawn from the femoral catheter, starting 5 s before the microsphere injection and continuing for 60 s.

The same procedure was repeated during the 0.5 g/kg glucose challenge and PET/MRI examination, with the exception that yellow microspheres were administered instead.

After the final PET/MRI examination, the pig was euthanized. Biopsies (1–2 g each) were taken from the different regions of the pancreas (six biopsies) as well as the right and left kidney cortex. The arterial samples and tissue biopsies were then frozen at − 20 °C until their microsphere contents were determined [2].

Each pig was positioned with the pancreas in the center of the scanner. 194 ± 46 MBq [¹⁵O]H₂O (corresponding to 6.1 ± 1.2 MBq/kg) was IV administered by a contrast pump (Medrad Stellant, Bayer, Whippany, NJ, USA) (0.8 mL/s for 10 s), followed by flushing with 15 ml NaCl (1 mL/s for 15 s). A dynamic PET sequence over 10 min was started simultaneously with the activation of the infusion pump.

The [¹⁵O]H₂O PET datasets were reconstructed into 26 frames, according to the following frame sequence: 1 × 10, 8 × 5, 4 × 10, 2 × 15, 3 × 20, 2 × 30, and 6 × 60 s.

Images were reconstructed using the VPFX algorithm (5 mm post-filter, diameter 50 cm). Vendor default MRI scan was used for attenuation correction.

[¹⁵O]H₂O PET images were analyzed in PMOD 3.7 (PMOD Technologies, Zurich, Switzerland). Regions of interests (ROIs) were segmented over the pancreas, spleen, and kidney cortex on transaxial co-registered MRI images. The ROIs were summed into volumes of interest (VOIs) and overlaid on transaxial PET SUV-corrected images. In the descending aorta, single voxels fully within the vessel lumen (to minimize the partial volume effects) were segmented directly on early PET SUV-corrected images.

For each PET examination, perfusion (expressed as mL/min/g tissue) was calculated by the 1-tissue compartment (1TC) model for [¹⁵O]H₂O (PKIN module, PMOD 3.7), based on the dynamic [¹⁵O]H₂O uptake in the pancreas, spleen, and kidney cortex using the aortic signal as the input function. Parametric perfusion images were generated by the PMOD pixel-wise modeling module using the [¹⁵O]H₂O 1TC, according to Alpert (PXMod, PMOD 3.7).

The intravoxel incoherent motion (IVIM) technique was used for perfusion fraction f assessment [13]. IVIM refers to the movement of water molecules due to diffusion and capillary perfusion. Dimensionless perfusion fraction f represents the fraction of a voxel volume occupied by the capillaries. The technique is based on the separation of true diffusion (D) from perfusion related “pseudo-diffusion” (D*) using a diffusion-weighted sequence with multiple b-values. Fat-suppressed and respiratory-triggered DW images were acquired by a single-shot spin-echo sequence with echo-planar read-out. The following parameters were used: TE 70 ms, FOV 350 × 350 mm², spatial resolution 1.4 × 1.4 × 6 mm³, interslice gap 0 mm, bandwidth per pixel (BW) 976 Hz, and number of scans 3. Diffusion-encoding gradients (b = 0, 20, 75, 150, 450, and 900 s/mm²) were sequentially applied along the three orthogonal directions. The maps of (isotropic) apparent diffusion coefficient (ADC) were automatically calculated by the scanner with mono-exponential fitting.

The same pancreas, spleen, and kidney cortex VOIs, as used for the PET analysis, were transferred to the DWI images. Additionally, a VOI outside of the pig was used to generate the background signal. The background values were subtracted from the VOI values for each b-value in each scan. Then, the output was plotted against the B-values and the resulting curve fitted to the bi-exponential equation given below, where x = b-values, f = capillary volume fraction (or fraction of fast diffusion), D = true diffusion (mm²/s), and P = fast diffusion (perfusion related diffusion) (mm²/s) using the software package GraphPad 6.0 (GraphPad Software, La Jolla, CA, USA).

The number of microspheres in the pancreas, islets, kidney cortex, and arterial reference sample was calculated by manual counting. For this purpose, an inverted microscope equipped with both bright- and dark-field illumination was used. More than 2000 microspheres were counted in each sample. The blood flow could then be calculated, according to the formula Q_org = Q_ref× N_org/N_ref, where Q_org is the organ blood flow (ml/min), Q_ref is the withdrawal rate of the reference sample, N_org is the number of microspheres present in the organ, and N_ref is the number of microspheres in the reference sample [2].

Experimental procedures were performed as planned, except for in pig 1, where only one glucose challenge (0.5 g/kg) could be performed due to a delay in the [¹⁵O]H₂O production (Table 1). All pigs were normoglycemic (4.1–5.7 mM) at baseline. Administration of 0.5 g/kg glucose by IV increased the B-glucose to 10–15 mM (mean 12.6 mM, range 7.6–16.3 mM), while IV administration of 0.1 g/kg glucose increased the B-glucose more modestly (mean 6.3 mM, range 3.3–9.9 mM).

Pancreas and kidney cortex exhibited strong perfusion on PET parametric blood flow maps of the abdomen (Fig. 1a, b). Representative corresponding anatomical MRI (Fig. 1c, d) and ADC maps (Fig. 1e, f) from the same subject are also shown. Pancreatic perfusion during basal conditions was 0.73 ± 0.22 mL/min/g as assessed by [¹⁵O]H₂O PET (Fig. 2a). Moreover, 0.5 g/kg of glucose, administered by IV, increased the perfusion in the pancreas (0.95 ± 0.18 mL/min/g, p < 0.01) but not in the kidney cortex or the spleen. This corresponded to an average increase in the pancreatic perfusion by 35% (p < 0.05) (Fig. 2b). Thereafter, IV administration of 0.1 g/kg of glucose did not affect the pancreatic perfusion appreciably.

Capillary volume fraction (f), as measured by DWI-MRI and used as an index of perfusion, was 0.16 ± 0.05 in the pancreas (Fig. 2c). Administration of 0.5 g/kg of glucose by IV caused an increase in the pancreatic signal to 0.39 ± 0.10, corresponding to an average increase of 157% (Fig. 2d). No change was seen in the perfusion as measured by DWI-MRI, after the IV administration of 0.1 g/kg of glucose. The IV glucose challenge did not affect perfusion in the kidney cortex or the spleen as assessed by DWI-MRI. Pancreatic perfusion as measured by perfusion fraction f correlated with the assessment with [¹⁵O]H₂O PET (r = 0.81, R² = 0.65, p < 0.01) (Fig. 2e).

The blood flow value in the right and the left kidney cortex of pig 4 was in the range of 10%, verifying an adequate mixing of the microspheres with the arterial circulation. The blood flow in the exocrine pancreas as assessed by the microsphere technique (an average of six biopsies) was 0.7 ml/min/g at baseline, which increased by 29% to 0.9 ml/min/g after IV administration of 0.5 g/kg of glucose. In the same animal, the baseline perfusion assessed by [¹⁵O]H₂O PET was 0.8 ml/min/g, which increased by 23% to 0.98 ml/min/g after IV administration of 0.5 g/kg glucose. The islet specific blood flow, as determined by the microspheres, was 52 µl/min/g pancreas at baseline and 78 µl/min/g pancreas (an increase of 50%) after IV administration of 0.5 g/kg of glucose.