Ethan Nguyen, Javier Carmona, Arisa Matsuzaki, Naoki Kaneko, Katsushi Arisaka
10 min
Abstract
Introduction: Mechanical thrombectomy can cause vessel deformation and procedure-related injury. Benchtop models are widely used for device testing, but time-resolved, full-field 3D vessel-motion measurements remain limited. Methods: We developed a nine-camera, low-cost multi-view workflow for benchtop thrombectomy in silicone middle cerebral artery phantoms (2160p, 20 fps). Multi-view videos were calibrated, segmented, and reconstructed with 4D Gaussian Splatting. Reconstructed point clouds were converted to fixed-connectivity edge graphs for region-of-interest (ROI) displacement tracking and a relative surface-based stress proxy. Stress-proxy values were derived from edge stretch using a Neo-Hookean mapping and reported as comparative surface metrics. A synthetic Blender pipeline with known deformation provided geometric and temporal validation. Results: In synthetic bulk translation, the stress proxy remained near zero for most edges (median $\approx$ 0 MPa; 90th percentile 0.028 MPa), with sparse outliers. In synthetic pulling (1-5 mm), reconstruction showed close geometric and temporal agreement with ground truth, with symmetric Chamfer distance of 1.714-1.815 mm and precision of 0.964-0.972 at $τ= 1$ mm. In preliminary benchtop comparative trials (one trial per condition), cervical aspiration catheter placement showed higher max-median ROI displacement and stress-proxy values than internal carotid artery terminus placement. Conclusion: The proposed protocol provides standardized, time-resolved surface kinematics and comparative relative displacement and stress proxy measurements for thrombectomy benchtop studies. The framework supports condition-to-condition comparisons and methods validation, while remaining distinct from absolute wall-stress estimation. Implementation code and example data are available at https://ethanuser.github.io/vessel4D
Alex: Wait—fixed vertices? So the connections don't change, even as the vessel moves?
Sam: Yes, *fixed* connectivity is key for consistent metrics. From the initial positions, they build a 3D neighbor net—connecting each point to its closest ones—then prune long links to avoid nonsense spans. This static edge graph lets them compute stretch on the same bonds frame-to-frame, distinguishing true deformation from whole-object drift.
Alex: Huh. So edge stretches become a stand-in for stress, revealing risks like at artery branches.
Sam: Precisely. They pull displacement in millimeters and a stress proxy in megapascals from regions of interest, highlighting where catheter paths strain vessels most. This full-field view uncovers patterns force sensors miss. The paper suggests it's a meaningful advance for benchtop testing.
Alex: So they get these edge stretches as a proxy for stress in specific spots. But how do they smooth out the noise from reconstruction to make those measurements reliable?
Sam: After tracking the fixed points frame by frame, small jitters appear from camera limits or lighting changes—like tiny shakes in a video that aren't real movement. To clean that, they average each point's shift with its neighbors, but smartly: close matches get full say, while big outliers get less weight, so local bends stay true. This repeats a few times per frame. It's graph-Laplacian filtering with Huber weights—the paper shows it cuts errors without blurring real deformation.
Alex: Okay, that quiets the wiggles while keeping the important stretches. Then for those regions of interest, like artery branches, what exact numbers do they pull out?
Sam: Regions of interest—or ROIs—are patches of connected points on the fixed graph, picked at landmarks like where arteries split. For each frame, they take the median shift distance across points in the ROI, then find the peak value over the whole procedure—that's max-median displacement in millimeters. For stress, they measure how much each edge lengthens from start, divide to get a stretch ratio, then plug into a simple formula from rubber-like material physics: stress grows with the square of stretch minus a shrink term. Using the silicone's properties gives megapascals as a comparable load gauge.
Alex: Huh—so median avoids outliers, and that formula turns raw pull into something like wall pressure. Does the smoothing make a clear difference in validation?
Sam: Yes, the study ablates it on synthetic tests—reconstruction matches ground truth much closer with filtering, per point-set distances and consistency checks. Without it, noise inflates errors, but this keeps deformation gradients sharp. Overall, it enables reliable ROI summaries. The evidence points to a solid, low-cost tool for safer thrombectomy design.
Alex: So the filtering sharpens those ROI metrics. But before trusting benchtop results, they test on fake data with known answers, right?
Sam: Yes, they create synthetic tests in a 3D graphics program called Blender, rendering a vessel model that moves in exact ways—like sliding the whole thing without bending, or pulling just one spot. This checks if the reconstruction and analysis match the known truth.
Alex: Okay, rigid slide first—why that?
Sam: A rigid slide moves every point together, like carrying a balloon without squeezing it—no real stretching happens. The system should detect near-zero stress everywhere, proving it ignores whole-object motion and only flags true deformation. The study shows median stress near zero megapascals, with regional displacement error at 0.091 millimeters—a clear match to ground truth.
Alex: Huh, so it separates drift from stretch properly. And for the pulling test?
Sam: To compare point clouds, they use Chamfer Distance—like checking how well two handfuls of confetti overlap from both sides. Across pulls, it stays low. Agreement is strong: reconstructed max-median shifts line up almost perfectly with truth, slope near 1 and R-squared at 0.992. Stress bias is just 0.004 megapascals. Without filtering, stress agreement loosens, confirming its value. The paper suggests this validates the workflow for benchtop use.
Alex: Okay, so the synthetics check out closely. With validation in place, how does this play out on actual benchtop thrombectomy tests?
Sam: They apply the same workflow to real phantom trials, comparing two common catheter paths into the brain artery model. One starts from the neck section of the carotid artery—called cervical access—threading farther to reach the clot. The other enters near the artery's end—terminal access—taking a shorter route. The study runs one trial each, pulling the clot out while tracking full-surface changes.
Alex: Right, shorter route might mean less pulling on branches. What patterns emerge at those key spots like the MCA bifurcation?
Sam: Displacement fields spread wider and peak higher with cervical access, especially around the middle cerebral artery split—or MCA bifurcation—where vessels branch. The max-median shift there hits 3.144 millimeters versus 0.846 for terminal—a roughly 3.7-fold difference. Stress proxies follow suit, about 1.8 times higher at the bifurcation. This shows cervical paths concentrate more strain exactly where perforations risk climbing, patterns force sensors at tips overlook.
Alex: Huh—so fuller tracking flags route-specific hotspots invisible before. Since it's single trials, they frame it carefully?
Sam: Yes, as directional observations from matched conditions—not statistical proofs, given the scope. Videos confirm the visuals: broader deformation plumes with cervical entry. The paper positions this as a standardized, low-cost way to compare maneuvers, guiding toward terminal access or gentler designs to cut injury odds in stroke care. Evidence suggests it's a meaningful tool for benchtop refinement.
Alex: So these directional differences—like higher strain at the MCA bifurcation with cervical access—align with what prior studies suggest about access strategies and injury risks.
Sam: Yes, the paper notes consistency with earlier work showing that catheter position and techniques affect vessel loading. It highlights three key outcomes: near-zero stress in rigid-motion tests, good agreement on synthetic pulls for task-relevant metrics, and clear trends between the two benchtop paths in displacement and stress at ROIs.
Alex: Okay, so it's a starting point—reliable for comparisons within its setup, but needs expansion for real-world claims. What are the limitations?
Sam: They present it as a demonstration for comparison, not statistical inference—meant to show the method's potential in spotting patterns. Limitations include the single-trial design, so no broad averages; the synthetic tests skip real-video segmentation challenges like cluttered backgrounds; and the stress proxy measures surface loading relatively, not absolute wall stress through the vessel thickness or inside pressures. Silicone phantoms help control variables but don't match living tissue behaviors fully.
Alex: Huh, so it gives labs practical tools—like those ROI summaries—to test safer paths without guessing on hidden strains.
Sam: The evidence supports it as a coherent framework for hypothesis generation in device and strategy testing, separating kinematics from full biomechanics. That's the core contribution here—a standardized way to quantify regional risks in benchtop work.
Alex: Makes sense... a solid step for improving stroke procedures step by step. Thanks for listening to ResearchPod.