⚡ electrical engineering

Contact-based microrobotic perception for vascular reconstruction

This paper presents a safe and precise method for early vascular diagnosis using anisotropic, magnetically controlled microrobots that navigate blood vessels and reconstruct 3D lumen topology in real-time by emitting detectable magnetic signals, thereby overcoming the limitations of traditional angiography.

Original authors: Ziyang Zhang, Xinyi Wu, Haoyu Liu, Jiacheng Ma, Tingting Yu

Published 2026-07-13

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Ziyang Zhang, Xinyi Wu, Haoyu Liu, Jiacheng Ma, Tingting Yu

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine trying to map a dark, winding cave system, but you aren't allowed to bring a flashlight, a camera, or even a flashlight. In fact, the only way to see the cave is to bump into its walls with your nose and remember exactly where you are. That is essentially what this team of researchers from South China University of Technology and Shanghai Jiao Tong University has done, but instead of a cave, they are mapping tiny blood vessels, and instead of a nose, they are using a microscopic robot.

The Problem with the Old Way
Usually, doctors look inside blood vessels using "angiography," which is like taking an X-ray movie of your insides. But this has some big downsides: it uses radiation (like X-rays), it requires injecting a chemical dye that can be toxic to kidneys, and it sometimes misses tiny, early-stage blockages because the resolution isn't sharp enough. The authors argue that relying on these external cameras isn't the best way to find small problems early. They want a method that doesn't need radiation, doesn't need toxic dyes, and can see the tiny details.

The Solution: A Tiny, Magnetic Explorer
Instead of looking at the robot from the outside, the team made the robot itself the sensor. They built a microscopic robot (about the size of a grain of sand, specifically 230 to 310 micrometers) out of a clear, hard plastic resin embedded with tiny magnetic particles.

Think of this robot as a tiny, magnetic marble. Because of how they built it, it has a "magnetic personality" that always points in a specific direction relative to its shape. When the scientists outside the body use a special set of five giant electromagnets (coils) to spin and push the robot, the robot moves. But here's the magic trick: as the robot moves, it sends out a magnetic signal. A grid of 16 tiny sensors sitting under the patient (or a model) catches this signal.

By listening to the magnetic "whisper" of the robot, the computer can figure out exactly where the robot is in 3D space and which way it is facing, with an accuracy of about 58 micrometers (that's less than the width of a human hair). The robot doesn't need a camera; it just needs to touch the wall. When it bumps into a vessel wall, it stops or changes direction. By recording every bump and turn, the computer builds a 3D map of the vessel, point by point.

Two Shapes for Two Jobs
The team didn't just make one robot; they made two, like a Swiss Army knife with two different tools:

The Cube: This robot is great for speed. It can tumble and roll quickly to map out the big picture of the whole network, finding where the branches split (bifurcations). It's like a scout running ahead to find the main roads.
The Disk: This robot is flat and stable. It's designed to roll along the bottom of the vessel and take very precise measurements of the wall's shape. It's like a surveyor carefully measuring the width of a road to see if it's getting too narrow (stenosis) or too wide (aneurysm).

How They Stayed in Control
Moving a tiny robot inside a flowing river of blood is hard. If the blood flows too fast, it washes the robot away. The team found that if the robot stands up straight, it catches the most current and gets swept away. But, they discovered a clever "zigzag" dance. By making the robot move at a slight angle (about 15 degrees) rather than straight against the flow, it could resist water speeds up to 13.8 cm/s without getting washed away.

They also invented a "Magnetic Moment Locking" technique. Imagine trying to push a spinning top; if you push it while it's wobbling, it goes flying. The team figured out how to "lock" the robot's magnetic orientation first, then gently guide it to the next spot. This stopped the robot from jumping around uncontrollably, ensuring the map they built was smooth and accurate.

The Results: A Map Without the Camera
In their tests, they used plastic models of blood vessels (phantoms) with tubes ranging from 0.8 mm to 4 mm wide. They guided the robot through these tubes without ever seeing the tubes with a camera.

The Map: The robot successfully mapped complex networks with two splits and three branches.
The Accuracy: The 3D model they built was incredibly close to the real plastic model. The center line of the vessel was off by only about 58 micrometers (a relative error of 5.7%), and the estimated width of the vessels was off by only 4.7%.
The Speed: It took about 20 to 23 minutes to map a small section of the network.

What This Means (and What It Doesn't)
The authors are very clear about what they have achieved and what is still ahead. They have demonstrated in a lab setting that this "contact-based" method works to build a 3D map of blood vessels without radiation or dye. They have measured the accuracy to be around 5% error in these plastic models.

However, they explicitly state that this is not yet a clinical tool for humans. They argue that the current method is slower than traditional angiography, so it's not ready for emergency situations where speed is everything. They also note that the current setup requires a fixed set of sensors and coils, which isn't practical for a moving human body yet. They suggest that for this to work in real hospitals, the sensors and magnets would need to be mounted on a robotic arm that can move around the patient.

In short, this paper suggests a new, safe, and radiation-free way to "feel" the inside of blood vessels using a tiny magnetic explorer. While it's not a magic wand for immediate surgery, it establishes a technical foundation for future diagnosis, proving that we can build a high-resolution map of our veins just by listening to a tiny robot bump its way through them.

Technical Summary: Contact-based Microrobotic Perception for Vascular Reconstruction

Problem Statement

High-resolution 3D imaging of small vascular blockages and lesions is critical for early cardiovascular disease prediction. However, current diagnostic modalities face significant limitations:

Digital Subtraction Angiography (DSA): While the clinical "gold standard," it is invasive, involves ionizing radiation, and relies on toxic iodinated contrast agents, limiting its use for routine screening or long-term monitoring.
Non-invasive Imaging (CTA/MRA): These methods often lack the spatiotemporal resolution required to resolve distal microvessels with slow flow or complex malformations, leading to undetected early-stage lesions.
Existing Microrobotics: Previous research has largely utilized microrobots as passive carriers for drugs or contrast agents. These systems typically rely on external imaging (X-ray, MRI, optical) for tracking, which reintroduces radiation risks or is limited by tissue opacity. They lack inherent environmental sensing capabilities, preventing true autonomous navigation.

There is an urgent need for a technology capable of reconstructing vascular networks at sub-millimeter accuracy without radiation exposure or toxic contrast agents.

Methodology

The authors propose a paradigm shift from "image-based tracking" to "contact-based robotic perception." Instead of visualizing the vessel, a single-body anisotropic microrobot acts as an in vivo probe, exploring the environment through physical contact and proprioceptive feedback.

1. Anisotropic Microrobot Design

Fabrication: Microrobots are fabricated from a photocured resin matrix embedded with vertically aligned NdFeB microparticles. A patterned-magnetization approach using UV lithography ensures uniform chain orientation.
Geometry: Two distinct shapes are engineered for specific functions:
- Disk-shaped robots (~310 μm diameter): Designed for stable surface conformity and precise sampling of vessel walls during localized reconstruction.
- Cubic robots (~230 μm side): Designed for enhanced mobility to rapidly navigate and map global network topology.
Actuation: A five-coil electromagnetic system (two orthogonal horizontal pairs and one vertical coil) generates rotating magnetic fields for rolling motion and vertical gradients for force control.

2. Magnetic Sensing and Localization

Sensor Array: A 4×4 planar array of 16 tri-axial Tunnel Magnetoresistance (TMR) sensors (10 nT resolution) detects the robot's magnetic field.
5-DoF Localization: The robot is modeled as a magnetic dipole. Due to rotational symmetry, a 5-degree-of-freedom (5-DoF) algorithm determines the 3D position and 2D orientation.
Algorithm: A hybrid approach combines a Deep Neural Network (DNN) for robust global initialization with a Levenberg-Marquardt (LM) algorithm using Variable Projection for high-precision local refinement. This achieves a computational latency of ~1 ms, enabling 160 Hz real-time tracking.

3. Hierarchical Reconstruction Strategy

The system employs a "coarse-to-fine" approach:

Global Topology Mapping (Cubic Robot): The robot rolls along the vessel bottom. Trajectory endpoints form a sparse point cloud. A topology extraction algorithm (k-NN, Minimum Spanning Tree, branch pruning, and B-spline fitting) reconstructs the network skeleton and identifies bifurcations.
Fine Geometric Reconstruction (Disk Robot): For specific segments, the disk robot performs radial sweeps. Assuming a circular cross-section, the system treats the reconstruction as an inverse problem. It minimizes the Euclidean distance between the measured point cloud and a modeled surface (defined by a centerline and radius function) using non-linear least-squares optimization.
Geometric Correction: Algorithms correct for the robot's physical dimensions and sampling bias (sampling only the lower arc of the vessel) to estimate the true vessel centerline and radius.

4. Maneuverability Enhancements

Zigzag Gait: To resist fluid drag, the disk robot utilizes a zigzag path, reducing the projected area against the flow.
Magnetic Moment Locking (MML): A "Lock-Project-Actuate" sequence stabilizes the robot's pose before actuation, preventing angular mismatches and uncontrolled jumps when navigating slopes or non-planar environments.

Key Results

Localization Accuracy: In vitro tests demonstrated effective tracking depths ( $z_{max}$ ) of approximately 5.5 mm (disk) and 6.0 mm (cube) with an error threshold of 200 μm.
Topology Mapping: In a vascular phantom with bifurcations and branches of 0.8–1.2 mm, the system successfully mapped the network connectivity. The reconstructed centerline showed a mean deviation of 58 μm (5.7% relative error) against the ground truth.
Fine Reconstruction: In a curved phantom with diameters varying from 2 to 4 mm, the system achieved a centerline deviation of 155 μm (5.4% relative error) and accurately reconstructed the radius profile.
Flow Resistance: The zigzag gait allowed the disk robot to withstand flow velocities up to 13.8 cm/s at a 15° heading angle, significantly outperforming standard straight rolling.
Navigation: The system successfully guided the robot through a blinded maze and vascular network using only real-time magnetic feedback, without any external imaging.

Significance and Claims

The paper claims to provide a safe, real-time, and highly precise solution for early vascular diagnosis that is independent of external imaging hardware.

Radiation-Free and Non-Toxic: By eliminating the need for ionizing radiation and contrast agents, the method addresses critical safety concerns associated with traditional angiography.
Active Sensing: The transition from passive carriers to active, contact-based sensors allows for true environmental perception and autonomous navigation capabilities previously unattainable in microrobotics.
Clinical Potential: While the authors acknowledge that the data acquisition rate is slower than conventional angiography (limiting immediate acute clinical translation), they posit that the system is ideal for scenarios requiring uncompromising safety, such as routine cardiovascular health screening and long-term progressive monitoring of chronic lesions.
Retrievability: Unlike magnetic particle swarms, the single-body microrobot is a retrievable probe, eliminating the risk of foreign material retention or embolization.

The work establishes a technical foundation for future radiation-free in vivo cardiovascular diagnosis and surgical navigation, demonstrating that anisotropic microrobots can serve as intelligent diagnostic devices capable of high-fidelity 3D vascular reconstruction.