Arterial Model
The rigid planar arterial model, fabricated using stereolithography processes, included left and right gastric, proper hepatic, gastroduodenal, and splenic arteries with flow terminating into 13 compliant distal vessels (Fig. 2). Central to the design of this model was replicating the hemodynamics (pressures, flow rates, pulsatile flow characteristics) and anatomical geometry (vessels diameters) both proximal and distal to the embolic injection point. For bariatric embolization procedures, embolic microspheres are infused from a delivery catheter at the left gastric artery, where they proceed towards the right gastric artery and small arterial branches supplying the lesser curvature of the stomach, ideally implanting in the latter. Flow rates, vessel diameters, and pressures vary substantially from the injection point to the target location. The model geometry was based on 3D CT imaging and anatomical measurements of vessel diameters. Specifically, the left and right gastric arteries formed a collateral branch with two esophageal and ten gastric vessels (1.0 mm ID, 2.0 mm OD clear PVC tubing, labeled 1-12); the proper hepatic artery was comprised of a single branch (4.1 mm ID tubing, labeled 13). A pressure sensor was placed in the gastroduodenal artery for systemic pressure measurements, and pinch valves (located at each of the distal vessels) were included to regulate local pressures and flow rates. For microsphere collection and subsequent quantification, flow from each of these vessels was routed through laser-cut nylon mesh filters (100 µm filter rating) adhered to the base of nominal ½ inch PVC pipe couplings (Fig. 2b). Custom 3D-printed ABS brackets suspended the surgical tubing above the filter assemblies to visualization flow and verify embolization for each vessel (Fig. 2b).
A closed-loop, dynamically pressurized surrogate arterial system was assembled to facilitate experimental testing (Fig. 7). A custom-fabricated positive displacement pump and gear pump, connected in parallel, provided pulsatile flow with pressure profiles replicating those measured clinically (varying from 58 mmHg to 134 mmHg). The gear pump maintains the steady-state (time-invariant) component of the pressure waveform, while the positive displacement pump provides the pulsatile (time varying) component. At the proximal end of the arterial model, a fluid supply reservoir was connected to two computer-controlled pumps configured in parallel: a custom-fabricated positive displacement pump and gear pump (Greylor Corporation, Cape Coral, FL). Pinch valves at each of the 13 distal vessels allowed for precise adjustment of fluid resistance at each vessel. Real-time mass measurements of flow through the filters permitted the calculation of flow rates to both the proper hepatic artery and left and right gastric branches. Pumps were connected to the collection reservoirs to intermittently recirculate fluid back to the supply reservoir. One-way valves at the pumps prevented fluid leakage from the collection reservoirs between intermittent pump operations, ensuring accurate flow rate measurements. Flow through the proper hepatic artery was based on published right hepatic flow measurements [5-9] and the fraction of flow received by the right hepatic artery (RHA), 0.60 [8], yielding hepatic artery (HA) flow ranges from 48.3 to 375.0 mL/min. From this range, an experimental flow rate of 160 mL/min was chosen. A separate literature review was needed to determine physiological flow rates for the left and right gastric arteries [10-14]. Due to the lack of clinical measurements for right and left gastric arterial flow in humans, published arterial flow rate data for dogs was used to determine a ratio between left gastric and proper hepatic arterial flow (Table 3). The target arterial flow rate value (160 mL/min for the nominal case) was then divided by this ratio (2.84) to determine total flow to the gastric branches (56.3 mL/min).
Since flow through the LGA is known to be significantly higher than that of the RGA, and since clinical RGA flow rates are not readily available in the literature, only LGA flow was considered in this study. Flow per branch through each of the esophageal branches was set equal to the average flow per gastric branch (5.6 mL/min). Because whole blood exhibits non-Newtonian characteristics at physiological flow rates, its viscosity varies widely (by nearly an order of magnitude) throughout the cardiac cycle [15,16]. A study by Lowe et al. reported a mean blood viscosity (corrected for hematocrit level) of 3.49 cP at systolic shear rates (measured ex vivo) [17]. To achieve mean blood viscosity (corrected for hematocrit level) [17], a 25/75 glycerin/water solution was selected as the model’s working fluid. Measurements taken using a HAAKE™ Viscotester™ 550 (ThermoFisher Scientific, Waltham, MA) confirmed a viscosity of 3.48±0.42 cP at a shear rate of 150 s-1.
Table 3. Published left gastric and hepatic arterial flow rates for dogs
Artery
|
Investigator(s)
|
Average Flow (mL/min)
|
Average Dog Weight (kg)
|
Average Flow per Dog Mass (mL/min/kg)
|
Average for Anatomy (mL/min/kg)
|
Ratio, HA/LGA
|
Left Gastric
|
Jacobson
|
39.0
|
14.0
|
2.79
|
2.70
|
2.84
|
Naruse
|
33.9
|
13.0
|
2.61
|
Hepatic
|
Burton-Opitz
|
143.5*
|
19.1†
|
7.8* (7.5 by authors’ calculation)
|
7.65
|
Grodins, et al.
|
144.0*
|
19.1‡
|
7.5*
|
*reported in [12], †from [13], ‡from [14]
Microsphere Injection
The four gastric vessels immediately distal to the injection site (labeled 3 through 6 in Fig. 2) were targeted for embolization. Each run included an initial injection of 1.5 mL of a BeadBlock (BTG International Ltd., London, UK) mixture prepared by diluting BeadBlock solution (300-500 µm diameter for all tests of this study) from the manufacturer’s syringe with radiographic contrast and saline. Contrast was added at a 1:1 contrast:BeadBlock ratio and saline at a 2:1 saline:BeadBlock ratio. Each injection was followed by a saline flush, also of 1.5 mL volume. Preliminary test runs indicated that a dosage of 1.5 mL with the 3:1 dilution ratio provided an adequate number of microspheres to embolize the target vessels. This volume also matched the nominal pre-programmed dosage of the Endobar manifold.
Tests were repeated using three catheters: a standard micro-catheter (Renegade HI-FLO 2.8 Fr, Boston Scientific, Marlborough, MA), a Surefire anti-reflux catheter (SHF-38120-mT, TriSalus Life Sciences, Westminster, CO, formerly Surefire Medical), and an Endobar occlusion balloon catheter (Endobar Solutions, Orangeburg, NY). The latter is a prototype device, designed specifically for bariatric embolization. As of this publication, it has undergone clinical trials and is in the process of FDA approval. While the standard catheter has no anti-reflux features, the Surefire catheter includes an expandable semi-permeable tip; the Endobar catheter features a deployable balloon just proximal to its tip (Fig. 1).
For all three catheter types, manual injections, computer-controlled injections with a constant rate of 1.5 mL/min, and computer-controlled versions with a pre-programmed injection profile were performed. For the Endobar catheter, a fourth injection method was employed: a custom manifold developed specifically for LGA embolization with the Endobar catheter. Thus, the design consisted of a 3 x 3 +1 statistical layout. All manual injections were performed by a single investigator (SJ) using a 6-mL syringe (Merit Medallion, Merit Medical, South Jordan, UT) and estimating dosage based on visual observation of the graduated marks. Due to their nature, the manual injections varied in rate. Post-test analysis of video found these to range from 5.25 to 10.80 mL/min.
Computer-controlled injections (constant rate and pre-programmed injection profile) were performed with a custom-made syringe pump operated through a LabView interface; manual agitation of the delivery syringe, immediately before injection, was required to promote suspension of the particles. The pre-programmed injection profile was devised to replicate manual delivery by a clinician. Prior to these studies, pressure measurements were dynamically recorded at the proximal end of the catheter during manual delivery of 1.5 mL of BeadBlock solution into the statically pressurized arterial model (Fig. 7, blue profile). A clinician performed the injection through an end-hole catheter. An injection profile (displacement of the syringe plunger over time) was then devised (Fig. 8, red profile), through an iterative process, to replicate the injection pressure profile. Finally, syringe displacements were scaled within the profile to provide 1.5 mL dosages, in accordance with other delivery methods.
The custom manifold, used exclusively for the Endobar catheter, included an internal syringe and automated valves for de-airing the infusion lines, agitating the embolic solution (for particle suspension), and delivering the embolic solution at a pre-programmed dosage and injection rate. While designed to deliver a 1.5 mL dosage at 1.5 mL/min, measurements found the average dosage and injection rates to be 1.22 mL and 1.22 mL/min, respectively, for the conditions of this study. To deliver an equivalent volume of BeadBlock solution, the ratio of BeadBlock was increased with respect to contrast and saline (volume ratios for manifold injections: 30.5% BeadBlock solution, 23.2% contrast, 46.3% saline; ratio for all other injections: 25% BeadBlock solution, 25% contrast, 50% saline).
Microsphere Quantification
Following injection procedures, each filter was allowed to air dry for increased microsphere opacity. Photographs of each filter were captured using a DSLR camera (Nikon D500 with 85mm AFS Micro NIKKOR lens, Nikon Corporation, Tokyo, Japan). Filters were backlit with the use of LED lights and an optical diffuser. A custom MATLAB application, developed through extensive adaptations of an open-source application (“FindCirclesGUI” by Brett Shoelson [18]), was used to automate the counting of microspheres in each image (Fig. 9). While most microspheres deposited on a single layer, microspheres depositing in multiple layers were distinguished by their darker color in relation to adjacent microspheres. In these cases, quantities of visible microspheres were doubled to account for stacking.