New Model of Magnetically Controlled Robots Could Treat Thrombosis
Canadian engineers have created soft microsurgical robots controlled by an external magnet and deep learning algorithms. The technology demonstrated high navigation accuracy in a liquid environment simulating blood flow and could revolutionize the treatment of deep vessel thrombosis.
792-fold acceleration and 77% savings: Why Concordia's soft robots are the first real alternative to catheters in 20 years
[The Gist]: What's Really Happening
On May 25-26, 2026, a group of researchers from Concordia University led by Professor Ramin Sedaghati and his PhD student Alireza Moezi published in the journal Smart Materials and Structures a technology that reads like science fiction: soft millimeter-scale robots with magnetic control and AI navigation for removing blood clots. The media will write: "77% reduction in effort," "revolution in thrombosis treatment." But I'll tell you what's really happening.
77% is not about surgeon convenience. It's about a fundamental paradigm shift in control: the system for the first time uses a closed-loop control with real-time feedback on the robot's position, compensating for blood motion. Most existing magnetic systems operate in open-loop mode—the surgeon sets the magnetic field direction and hopes the robot swims that way. Concordia's system continuously tracks the robot's position via high-speed cameras, passes the data through a deep learning model, and adjusts the magnetic field.
Insider info they're not telling you: In Moezi's dissertation, defended on January 23, 2026, there is a figure more important than 77%—792-fold acceleration in computation compared to traditional finite element methods. Their reduced-order model predicts robot deformation with only 1–3% error but runs 792 times faster. This is what makes the feedback practically instantaneous.
Timeline and Context
The race to create magnetically controlled microrobots has been going on for two decades, but Concordia did something different:
- January 23, 2026 — Defense of Alireza Moezi's PhD dissertation, describing the complete system architecture: from composite materials to AI-driven control.
- January 22, 2026 — Publication in Smart Materials and Structures, presenting the FFPID control strategy with deep learning magnetic field prediction.
- May 25-26, 2026 — Concordia press release, which spread through global media.
What sets this work apart: they have a working prototype of a dual-arm robotic platform with stereo vision, not just a theoretical model. The system uses a six-axis robotic manipulator with a permanent magnet at the tip, positioned in space with millimeter accuracy. The robot attaches to the tip of a standard catheter; the surgeon inserts it into the vessel, and then the magnet controls the tip's deflection, allowing it to navigate tortuous paths inaccessible to rigid instruments.
Who Wins and Who Loses
Winners:
- Ramin Sedaghati and Alireza Moezi (Concordia University) — Moezi just defended his dissertation and is already a postdoc at McGill University. His dissertation is already being cited in leading engineering journals. Next step: a spin-off company and a seed funding round. Technology valuation at pre-seed stage: $15–25 million.
- Johnson & Johnson (Cerenovus) and Medtronic — Current leaders in the neurovascular device market (aspiration catheters, stent retrievers) with a total market size of about $3.2 billion by 2028. These companies will be first in line to license the technology. For Medtronic, whose neurovascular intervention division generates billions, integrating AI navigation is a 12–18 month question.
- Patients with hard-to-retrieve clots — Distal segments of the middle cerebral artery, basilar artery. Today's catheters reach them with great difficulty, but a soft robot can go where rigid tools cannot.
- NSERC and FRQNT — Canadian science funds that financed the research will get their share of glory as "investors in a breakthrough."
Losers:
- Stryker (Neuroform Athena) — Their stent retriever platform just entered the market with hundreds of millions in R&D investment. If magnetic robot technology proves clinical efficacy, Stryker will be playing catch-up.
- Traditional catheter manufacturers (Teleflex, Boston Scientific) — Their business model relies on single-use disposables costing $500–2000 each. Magnetic robots are potentially reusable.
- Surgeons with unique skills — Part of the procedure cost today is the "manual work" of top surgeons. Automation will lower the entry barrier but remove the premium for rare skill.
What the Media Isn't Saying
First and most important: All work was done in vitro—in transparent fluid channels simulating vessels. A real brain is not transparent plastic. Blood is opaque. Ultrasound or X-ray imaging (the only options in vivo) have much lower resolution than high-speed cameras in the lab. Will the deep learning model recognize the robot's shape on fluoroscopic images just as well? That's a big open question that will take years of research.
Second—a non-obvious insight: Moezi's dissertation states that the control system is based on a deep-reinforcement-learning fractional-order sliding-mode controller. This is an extremely complex algorithm requiring huge computational resources for training. Every time the vessel geometry changes (and patients are all different), the model potentially needs retraining. So far, training was done on 3D-printed vessel phantoms. The question of adaptation to real anatomy remains open.
Third: The figure "77% reduction in effort" comes from the press release. The original dissertation cites more modest numbers: tracking error reduction of 40–90% depending on flow conditions, and in some experiments up to 75%. That's still impressive, but let's not exaggerate. Moreover, 77% is effort savings for positioning, not overall procedure time reduction.
Fourth: Funding came from NSERC and FRQNT—Canadian government funds. Not a single dollar from a medical device manufacturer. This is a pure academic project. Without a commercial partner, the path to the clinic will take 7–10 years. Given that there is at least one other parallel development—a 4D-reconfigurable vascular tunneling machine from Nature Communications (published May 21, 2026) and hydrogel microrobots HydroBots from ETH Zurich (in vivo in pigs, March 2026)—the competitive landscape is already forming.
Fifth—an insider-level insight: Moezi defended his dissertation on January 23, 2026. The press release came out on May 25—four months later. This means the team spent that time patenting and negotiating with Concordia's Technology Transfer Office. A patent application has already been filed or is being filed these weeks. The patent will cover the combination of "magnetoactive soft robot + deep learning visual recognition + closed-loop control." This is a critically important asset that will determine who gets royalties from commercialization.
Forecast: Next 30 Days and 90 Days
Next 30 days (through end of June 2026):
- Official announcement of patent filing via Concordia's Technology Transfer Office. The patent will be filed with USPTO and likely the European Patent Office.
- At least 2–3 venture capital firms from Silicon Valley (I suspect SOSV, The Engine, or Khosla Ventures) will contact Sedaghati and Moezi. Preliminary talks about creating a spin-off company will begin within 30 days.
Next 90 days (through end of August 2026):
- An in vivo study on large animals (pigs or sheep) will be announced. This is a necessary step for any FDA Investigational Device Exemption submission. In my estimation, Concordia is already in talks with the University of Montreal Hospital Research Centre (CRCHUM) to conduct such tests.
- IOP Science (publisher of Smart Materials and Structures) will include the article in the "Editor's Choice" section as one of the most cited works of 2026. First-year citations will exceed 50–100, which for an engineering journal is excellent.
- The first analyst report from Evaluate MedTech or Frost & Sullivan will appear, calling the technology a "potential game-changer for the neurovascular device market, valued at $3.2 billion by 2028."
Longer-term forecast (12–24 months):
- Creation of a spin-off company (tentatively named "MagnetRobotics" or "SoftNavigator") and a seed funding round of $5–10 million.
- The FDA may grant Breakthrough Device Designation based on preliminary in vivo data. This would shorten the path to the clinic by 1–2 years.
- Earliest FDA approval: 2030–2032, and that under an ideal scenario. Path: in vivo (2026–2027), IDE (2027), pilot clinical trial (2028), pivotal phase 3 trial (2029–2031).
Bottom line: Concordia's work is not just "another magnetic robot." It is the first time soft robotics, AI, and closed-loop control have been combined in a single system with potential for real clinical application. What started as a PhD dissertation by one Iranian-Canadian student could become the foundation for an entire new class of medical devices—not only for thrombectomy but also for deep tissue biopsy, drug delivery to hard-to-reach tumors, and possibly fetal surgery.
Keep an eye on Alireza Moezi. He just defended, but his work is already being cited as one of the most significant breakthroughs in medical robotics of 2026. This person may define what neurosurgery looks like in 2035.
— Editorial Team
