DYNWAY

Intelligent navigation for microsurgeryLesion boundary, instrument tip and safe distance stay visible throughout the resection

Microsurgical navigation for neurosurgery and microsurgery. It mounts onto the microscope already in the room and rests on two key innovations: sub-millimeter intraoperative real-time registration and micron-scale real-time instrument tracking.

Navigation has to keep answering “where am I”. Existing systems register once, before surgery, and fail the moment tissue deforms — continuous intraoperative registration remains unsolved.

  • 0.164mm initial registration error±0.059, max 0.270 · measured in live-mouse craniotomies
  • 0.009mm instrument tracking accuracy9 µm · fiber tracking point · bench measurement
  • 15/15in vivo craniotomies succeededMurine · Univ. of Tokyo ethics approval A2023M042-07 · the in vivo subset of 74 animal studies
  • 2029Q2 · NMPA Class III submission targetEarliest Q3 2028 · certification targeted 2030

The project remains in R&D and validation · Core technologies validated on the bench and in animal studies; end-to-end system navigation accuracy will be established by third-party testing after integrated prototype V1

The surgeon at the microscope; after coaxial splitting, the visible and infrared paths drive live imaging, instrument tracking and image registration — the left screen overlays the lesion boundary, the right screen shows 3D navigation
One light path split three ways — registration, tracking, imagingSystem concept

PROJECT FILM

The system in one video

A full walkthrough of the operating principle, the prototype, and the experimental setup (with narration).

Project introduction video · with narration1:12

Clinical Need

Microsurgery lacks a quantitative reference for deep anatomy

The relative positions of lesion boundaries, instrument tips, and critical anatomy are difficult to judge directly. Critical boundaries and residual tissue are often below 0.5 mm — beyond the resolving limit of the human eye and the depth of field of an operating microscope — while conventional navigation accuracy of 1–3 mm sits orders of magnitude away from the 20–25 µm scale of microsurgery. As procedures move deeper and toward smaller structures, reliance on individual experience increases, along with training time and variability in risk.

Relevant procedure volume
2.25million/year

Estimated annual volume of core microsurgical procedures in China: approximately 1.1545 million neurovascular, 512,100 malignant brain tumor related, 300,600 ENT, and 283,800 ophthalmic (excluding cataract).

Clinical value of resection extent
+29.4months

Reported mean survival benefit of gross total versus subtotal resection for grade IV glioma.

Neurosurgeon density
0.78per 100,000 people

Density is about 30% of that in developed countries, with a 10–15 year training cycle. In microvascular free-flap surgery, for example, novice anastomosis patency is only 60–70%, and roughly 600 cases are needed before the success-rate curve plateaus. Navigation supports decision-making and shortens the learning curve.

Procedure-volume data: China Health Statistics Yearbook. Other clinical figures are drawn from public research; definitions, reference years, and indication scope may vary between sources. The lack of deep-field and boundary information was confirmed in interviews with neurosurgeons at Sun Yat-sen University Cancer Center.

Limits of existing solutions: why millimeter-scale navigation cannot enter the microscopic field

Limits of Existing Solutions

Why millimeter-scale navigation cannot enter the microscopic field

Conventional navigation is designed around rigid bony fixation, large reflective markers, and an open field. Once microsurgical resection begins, all three premises break down at once — drift, scale, and features compound, which is why intraoperative dynamic registration that follows soft-tissue deformation remains the key unsolved technical bottleneck.

Lesion shifting during surgery, so the planned position no longer matches the actual one
01 · Preoperative and intraoperative positions divergeTissue drift
Intraoperative tissue drift
Effective only for preoperative planning

After the dura is opened, cortical shift can reach 1–2 cm (as reported in the literature), and every subsequent maneuver and resection step keeps changing tissue position. Existing navigation cannot re-register in real time, so the on-screen position gradually diverges from the actual surgical field.

A microsurgical forceps tip of about 0.3 mm beside an 11 mm reflective marker sphere: the sphere cannot be mounted on a microsurgical instrument
02 · Reflective sphere versus microsurgical instrument tipScale comparison
11 mm vs 0.3 mm
Physical scale mismatch

Existing navigation carries system error of roughly 1–3 mm, insufficient for microscale boundary judgment. Conventional reflective spheres are about 11 mm in diameter against 0.3 mm microsurgical instrument tips, and readily obstruct a narrow field.

Smooth, uniform brain surface model under the microscope, offering few stable features for an algorithm to lock onto
03 · Very few repeatable features in a field a few millimeters wideFeature scarcity
No features to rely on under deformation
Rigid registration and feature tracking fail together

Soft tissue shifts non-rigidly as soon as the dura is opened, so the assumptions behind rigid-body and ICP-type registration no longer hold. A microscopic field spans only a few millimeters and offers very few stable, repeatable features, and glare, bleeding, and occlusion further alter the local image.

Sources: items 01 and 02 reflect feedback from neurosurgeons at Sun Yat-sen University Cancer Center; the magnitude of brain shift is from the literature (Nimsky et al. 2000; Nabavi et al. 2001); other figures are from manufacturers' published specifications and the literature; item 03 is a conclusion from the team's own algorithm validation.

Product Concept

Keeping “where is the instrument” and “what is this tissue” true throughout resection

Navigation rests on two conditions: where the instrument is right now and what the tissue in front of it is. Both must hold throughout resection. A system that answers them once, before the incision, is providing a plan — not navigation.

MAP · EXISTING NEUROSURGICAL NAVIGATION

A preoperative map: registered once, never updated during surgery

  • Core capability: preoperative planning plus millimeter-scale probe localization during surgery
  • Where is the instrument — real time, but at millimeter scale: 1–3 mm system error, tracking reflective spheres about 11 mm across, against instrument tips on the 0.3 mm scale
  • What is this tissue — registered once before surgery and not updated after tissue shifts
  • Tissue keeps moving during resection, so coordinates cannot stay in sync: the guidance reflects the preoperative state, not the field in front of the surgeon
NAVIGATION · DYNWAY

Real-time intraoperative microsurgical navigation

  • Core capability: real-time sensing, dynamic registration, and quantitative cues
  • Where is the instrument — at micron scale: the fiber tracking point sits at the instrument tip
  • What is this tissue — dynamic intraoperative registration that updates with soft-tissue deformation
  • Keeps answering: where am I now | can I keep resecting | how much more can I safely remove
Objective: bring boundary, distance, and risk cues into the current view without disrupting the surgeon's core operating habits.Product value still requires continued validation in real clinical workflows, against defined baselines, and through surgeon feedback.

Key Technologies · Product Forms

One product, three forms, built on two key technical innovations

The two key innovations are sub-millimeter intraoperative real-time registration and micron-scale real-time instrument tracking. In the product they take three forms: a split-path stereo operating microscope and trackable microsurgical instruments carry the instrument tracking, and the 3D navigation software carries the registration. The result appears as boundary and distance cues inside the surgeon's existing microscopic view. Together they form a layer of intraoperative perception.

Product Forms

Where each of the three forms sits during an operation

The system mounts onto the operating microscope already in the room and leaves the operating-room layout unchanged. The three forms marked here, plus the overlay inside the surgical field, are all of it that is visible in the operating room. The image is a concept rendering, not a photograph.

Select a numbered point in the scene, or any form below, to mark its position and open the technology behind it, with its study records.

The complete system in an operating room: the 3D navigation interface showing the preoperative 3D planning model at left, the surgeon working at the split-path stereo operating microscope with trackable microsurgical instruments at center, and the live surgical field with an overlaid lesion boundary and safety margin on the right-hand screen
01HARDWARE · SPLIT-PATH IMAGING Split-path stereo operating microscope One optical path captures visible and near-infrared light OpenNow showing
02INSTRUMENT · FIBER TRACKING Trackable microsurgical instruments < 0.1 mm fibers embedded in forceps and scissors OpenNow showing
03SOFTWARE · REGISTRATION 3D navigation interface Initial registration aligns; dynamic registration follows deformation OpenNow showing

01 · HARDWARE · SPLIT-PATH IMAGING

Split-path stereo operating microscope: making the tracking point visible in the field

One optical path, two purposes. Coaxial beam splitting lets the same field capture visible and near-infrared light at once, with no filter-set switching: visible light renders the surgical field, and near-infrared picks up the signal from the fiber tracking points. The three images show the coaxial split path, the split-path module validation prototype, and integration with an operating microscope. The split-path module has completed bench validation.

  • Visible and near-infrared light share one optical path for simultaneous field visualization and tracking-signal acquisition.
  • Stereo vision provides the 3D measurement foundation, converting microscopic images into depth and relative-position information.
  • Once integrated with the microscope, the system supplies stable observations for instrument tracking and intraoperative registration.
0.11 mmSystem error across a 20 mm microscopic field (split-path module, bench validation)
Coaxial optical-path diagram for split-path stereoscopic microscopy
01 · Coaxial split optical pathVisible + near-infrared
Split-path module validation prototype
02 · Split-path module validation prototypeBench validation
Microscopic navigation prototype with integrated split-path stereo vision
03 · Microscope integrationSystem validation prototype

The microscope platform combines imaging and metrology, preserving the live surgical field while providing structural, depth, and relative-position references.

02 · INSTRUMENT · FIBER TRACKING

Building the tracking marker into the instrument tip

Optical fibers under 0.1 mm in diameter are embedded in forceps and scissors to form active tracking points at the tip, preserving the instrument's original dimensions, weight, and handling. Conventional reflective spheres are about 11 mm across, while a microsurgical instrument tip is on the 0.3 mm scale — so the tracking marker moves from a reflective sphere to a fiber at the tip, with tip position linked to the preoperative model.

< 0.1 mmFiber diameter
0.009 mmTracking-point localization accuracy (9 µm, point-by-point localization, bench measurement)
Real-time demonstration of a fiber probe navigating on a mouse skull
Fiber probe · mouse-skull navigationReal-time demonstration
Trackable microsurgical instrument with embedded ultrafine fibers
Trackable microsurgical instrumentStructural concept
3D display of tip motion with fibers embedded in forceps
Fibers embedded in forceps · 3D display of tip motionBench demonstration
Validation sample of forceps with embedded optical fibers
Fibers embedded in the instrumentValidation sample
Fiber forceps against a 3 mm scale bar, with the corresponding tip-coordinate readout
Against a 3 mm scaleBench photo
Hand-held fiber forceps, with the fibers routed along the body to the tip
Hand-held · fiber routingBench photo

The fiber-based active tracking point is covered by a patent application (Japanese Patent Application No. 2025-284200, accepted for filing), and proof of principle is complete; the precursor method, stereo keypoint 3D reconstruction, was published in IEEE T-ASE (2026).

03 · SOFTWARE · REGISTRATION

Sub-millimeter intraoperative real-time registration: align first, then follow deformation

Registration runs in two steps. Initial registration uses semantic features such as vessel course and bone-surface texture in place of manually placed markers, needs no implanted bony markers, and has been validated in live-mouse craniotomies (ethics approval A2023M042-07, the University of Tokyo). Dynamic registration selects landmarks automatically on the tissue surface within the microscopic field, keeps assessing their reliability, and follows soft-tissue deformation; this step is currently validated on public data and in simulation, with in vivo validation scheduled for the next stage.

0.164 mmInitial registration error (±0.059, max 0.270) · measured in live-mouse craniotomies
15 / 15In vivo mouse craniotomies succeeded, 0 failures
< 5 minTime for one initial registration · measured

STEP 1 · INITIAL REGISTRATION · MEASURED IN LIVE-MOUSE CRANIOTOMIES

Rigid registration from salient visual anatomical features

Semantic-feature constraints: vessel course and bone-surface texture replace manually placed markers, with no implanted bony markers. The table places three classes of registration method side by side and notes the source of each figure; soft-tissue deformation is handled by step 2, dynamic registration.

Registration methodRegistration errorMicroscopic deformationSource
Bone-anchored registrationapprox. 0.8–2 mmNot applicablePublished literature
Conventional ICP surface matching2.7–5.3 mmNot applicablePublished literature
This project · semantic registration0.164 ± 0.059 mm (max 0.270 mm)Handled by dynamic registrationMeasured in live-mouse craniotomies

The bone-anchored row is registration error; overall system error is typically 1–3 mm. Literature sources: intraoperative-CT registration, 0.86 mm; surface matching, up to 5.3 mm; ICP, 2.7 mm. Literature values and this project's measurements differ in experimental conditions, sample size, and statistical definitions and are not directly comparable; they are shown to indicate the difference in order of magnitude. This project's data come from live-mouse craniotomies, with a median translational registration error of 0.16 mm; the full statistical definitions and raw records are available in the diligence pack.

Robot performing a mouse craniotomy with navigation assistance
Robot-assisted mouse craniotomyHigh-precision operation demonstration
Automated registration and calibrationLive process
Close-up of a mouse cranial window
Cranial-window close-upApproximately 2 mm
Scale reference for a small cranial-window model
Small cranial-window modelScale reference
Mouse-head point cloud and coordinate reconstruction
Point cloud and spatial coordinates3D reconstruction

STEP 2 · DYNAMIC REGISTRATION · PUBLIC DATA AND SIMULATION

Following soft-tissue deformation and recovering after interruption

Dynamic landmarks are selected automatically on the tissue surface within the microscopic field, assessed continuously for reliability, and replaced as soon as they fail. Displacement of the deep lesion is inferred from the surface motion field, with no reference frame and no intraoperative imaging.

Grasp demonstration · virtual lesion model updates in real timeAlgorithm workflow
01
Dynamic landmarks

Selected automatically on the tissue surface, assessed continuously, and replaced on failure, keeping instrument trajectory, model, and live image in one spatial relationship.

02
Recovery after interruption

After the view is abruptly rotated by 90° / 180°, tissue-tracking accuracy is 54.0% / 58.7%, against 4.6%–6.6% for existing point-tracking models under the same conditions. Dataset: STIR, a public dataset of in vivo soft-tissue surgical video. The test corresponds to recovering after the microscope is repositioned or the field is re-entered.

03
Locating the lesion below the surface

The lesion that is not directly visible is inferred from surface motion, with localization error 75% lower than without deformation compensation. Baseline: pose left unchanged after initial registration; organ simulation.

FAIL-SAFE

Graceful degradation when features fail — never a silent correction

Bleeding, glare, and occlusion occur at any moment in the microscopic field. The system degrades in three steps according to registration confidence, and at no step does it keep issuing confident cues once the reference can no longer be trusted. The system advises and never acts; the decision always rests with the surgeon.

①
A few points fail · unnoticed by the surgeon

Outliers are rejected and the remaining landmarks maintain registration without interruption.

②
Solvable but low confidence · surgeon informed

The overlay freezes at the last trusted pose and the system prompts for re-registration.

③
Too few valid points to solve · surgeon takes over

The overlay is withdrawn and the surgeon decides whether to re-register.

In vivo validation of soft-tissue deformation is scheduled for the next stage: ex vivo porcine tissue first, to test adverse conditions such as bleeding and glare, followed by 6–12 in vivo porcine studies within 2027, with results forming the animal-study evidence for the registration submission. The navigation-system patent, Japanese Patent Application No. 2025-284200, has been accepted for filing; precursor methods were published in Advanced Robotics (2024) and IEEE RA-L (2026).

These results demonstrate R&D-stage system capabilities. Medical use, performance claims, and clinical approval status remain subject to subsequent validation and public regulatory documents.

04 · INTRAOPERATIVE OUTPUT

What the surgeon actually sees in the microscope

The intraoperative output is how form 03 presents itself: boundary, instrument tip, and distance all appear inside the existing microscopic view, so the surgeon does not shift gaze. Real-time navigation is linked to the preoperative plan, visualizing the tumor boundary and flagging critical structures. The system advises and never acts.

Concept of the intraoperative overlay: the live surgical field with a purple boundary overlay in the main view, the preoperative 3D model at top right, and the lesion segmentation view at bottom right
The purple outline is the tumor boundary given by the system, with the preoperative 3D model on the same screenConcept rendering
One frame of actual software output: the surgical field with the lesion contour at left, the 3D lesion model at right, and displacement and volume curves below
Surgical field with lesion contour (left) and 3D lesion model (right): one frame of the dynamic-landmark grasp demonstrationActual software output

Clinical Workflow

One added step, and no change to established surgical habits

The system mounts onto the existing operating microscope. Staff positions, instruments, and screen layout stay unchanged; the only step added across the whole procedure is one registration after the microscope is positioned. The surgeon receives boundary and distance cues inside the original microscopic view without shifting gaze.

Preoperative
Image import and planning

Import MRI/CT, segment the lesion and critical structures, and plan the approach.

Same as current practice
Intraop · Added
One non-invasive initial registration

Performed once after the microscope is positioned, with no implanted bony markers. This is the only step the system adds.

Under 5 min · measured
Intraop · Continuous
Automatic tracking of tissue change

The system continuously evaluates and updates the correspondence between the preoperative model and the live field, with no surgeon input.

Zero extra actions
Intraop · Decision
Boundary and distance cues

Presents lesion boundary, instrument tip, and relative position to critical structures inside the microscopic view.

In-field overlay
Postoperative
Procedure recording and review

Retains process data for case review, training, and accumulation of clinical evidence.

Added value

Sterility and instrument handling follow existing practice: the split-path module is covered with a single-use sterile drape without changing draping procedure, and fiber instruments are sterilized as conventional microsurgical instruments. Scrub staff add only drape placement and counting. Actual steps and per-stage timing are still being confirmed with collaborating clinicians.

Development and Intellectual Property

Development is progressing from prototype validation toward registration preparation

Proof of concept is complete, patents have been filed and granted, and seed-round investment commitments are in place. The product path runs through module prototypes, integrated prototypes, registration testing, clinical trials, and NMPA review. Timing remains dependent on experimental progress, regulatory consultation, and funding.

2021–23
Project R&D

Completed method innovation and the core technical route, advanced human-machine interaction validation and first-generation prototype development.

Completed
2026
Module prototype

Animal studies and validation of the split-path camera module prototype completed; engineering iteration of the navigation software and instrument-tracking system under way; the company, Dynway Technology Co., Ltd., was established this year.

In progress
2027
Integrated prototype and registration testing

Two integrated V1 prototypes; registration testing starts in Q3, and pre-submission consultation with the review authority in Q4 fixes the clinical evaluation pathway.

Planned
2028
Small-sample clinical trial

A supplementary small-sample clinical trial, alongside registration prototypes and quality-system audit preparation; research-grade modules and OEM collaboration bring first revenue.

Planned
2029 Q2
NMPA Class III submission accepted

Registration documentation submitted, entering technical review; Q3 2028 at the earliest.

Target
2030
Expected approval

Includes buffer for review deficiency responses. This target depends on regulatory consultation and trial progress; the final timeline will reflect actual development.

Target

Commercialization does not hinge on this milestone: the Training & Assessment mode, OEM modules, and software licences generate revenue before approval. The staged path is set out in the Market, Competition, Funding and Milestones section below.

The team has 73 papers in this field, 15 of them core papers (including in IEEE T-ASE and RA-L) and 11 with a team member as first author; the robotic autonomous mouse-craniotomy drilling system received the IROS 2023 Best Application Paper Award. 74 animal studies have been run to date under ethics approval. The research collaboration network spans four institutions in China and Japan: Sun Yat-sen University Cancer Center (Neurosurgery) and the University of Tokyo Faculty of Medicine (ophthalmology, pediatric surgery) have signed letters of intent for clinical validation; Zhongda Hospital, Southeast University, and Teikyo University School of Medicine have expressed intent for research collaboration.

Registration pathway: NMPA Class III, four gates from testing to acceptance
2027 Q3
Testing and standards conformity

Electrical safety, electromagnetic compatibility, software, and biological evaluation, against GB 9706.1, YY 9706.102, GB/T 25000.51, and GB/T 16886; testing is planned at a provincial medical-device testing institute.

Testing starts
2027 Q4–28
Clinical evaluation pathway

Pre-submission consultation with the review authority in Q4 2027 fixes the pathway, followed by a supplementary small-sample clinical trial with ethics approval and filing. Marketed surgical-navigation products serve as the comparable devices; because the core principle differs, the baseline plan is a single-arm trial with an accuracy endpoint at 2–3 centers.

Consultation · trial
2027 Q4–28
Registration units and quality system

Registration units and submission documents; manufacturing licence, quality-system audit, and supply-chain and process freeze. Two routes: contract manufacturing (asset-light, allows earlier acceptance) or an in-house assembly and test line; contract manufacturing is preferred.

System audit
2029 Q2
Submission accepted

Registration documentation submitted, entering technical review; Q3 2028 at the earliest. The statutory technical review is 90 working days, plus quality-system inspection and administrative approval; with deficiency responses the practical estimate is 12–24 months.

Target

One deficiency round is budgeted at 3–6 months and is already included in the 2030 approval target; quality-system documentation is prepared in parallel and the pre-submission consultation starts early. These dates are planning figures; the final timeline depends on regulatory consultation and trial progress.

How the accuracy was measured: calibration → measurement → statistics
01 · Calibration: a 3D calibration boardAn in-house 3D calibration board and performance-evaluation method for stereo microscopic imaging (patent application accepted for filing, CN202610805057.1), with the split-path camera module as the object calibrated. Calibration records cover board specification and feature spacing, number of repetitions, and residuals.
02 · Measurement: two kinds of metric, measured separatelySystem error of 0.11 mm across a 20 mm microscopic field; fiber tracking-point localization accuracy of 0.009 mm (9 µm). Measurement records cover sample size and repetitions, reported as RMS, maximum error, and the 95th percentile.
03 · Result: ablation of registration errorMedian translational registration error: ICP registration alone 8.92 mm; ICP + cranial-suture constraint 6.13 mm; ICP + contour constraint 25.37 mm; this method 0.16 mm. Overall system error of conventional optical navigation is typically 1–3 mm.

These are R&D-stage results from bench and live-mouse measurements. Raw calibration and measurement records are available in the diligence pack; end-to-end system navigation accuracy will be established by third-party testing after integrated prototype V1.

Animal-study detail: 15 live-mouse craniotomies
A laboratory mouse fixed in a stereotaxic frame after craniotomy
Craniotomy study setupThe University of Tokyo
Scale and outcome74 animal studies to date; of these, the 15 live-mouse craniotomies run after the system took shape succeeded 15/15, with 0 failures. Ethics approval A2023M042-07 (the University of Tokyo).
ConditionsMeasured mouse skull thickness 0.27–0.51 mm; bone windows of 2–6 mm in four shapes: circular, elliptical, square, and across a suture. Each procedure took 8.0–46.5 minutes, depending on window size and shape.
Success criteria and accuracySuccess required a fully formed bone window, intact underlying tissue, and no unintended damage to the skull. Contour machining error was about 0.04 ± 0.03 mm (quantitative experiment on a surrogate).
Next stageBench first: ex vivo porcine tissue to test adverse conditions such as bleeding and glare, then 6–12 in vivo porcine studies within 2027, with non-human primates assessed if clinical evaluation requires. Results form the animal-study evidence for the registration submission.

Study details are in the preprint arXiv:2609.12530 (submitted to IEEE T-RO).

Papers and sources: precursor methods for the key technologies are peer-reviewed and published
YearPaper topicJournal / conferenceTeam authorsRelates to
2026Bone micro-milling with automatic calibration and 3D surface fittingIEEE RA-LEnduo Zhao (first author), Xiaofeng LinCalibration and registration · ex vivo mouse 6/7
2026Microsurgical needle handling with real-time stereo keypoint trackingIEEE T-ASEXiaofeng LinInstrument tracking (the object is a suture needle)
2025Autonomous drilling of mouse cranial windowsIEEE T-ASEEnduo Zhao (first author)Validation platform · ex vivo mouse 14/20
2025Estimating the state of the drilled object through active contactIEEE RA-LXiaofeng Lin (first author), Enduo ZhaoValidation platform
2025Completion recognition in autonomous micro-drillingScientific ReportsEnduo Zhao (first author)Validation platform
2025A multimodal digital twin for micro-drillingIJCARSEnduo ZhaoValidation platform
2024Skull-feature registration with a microscopic stereo cameraAdvanced RoboticsXiaofeng Lin (first author)Registration · precursor method
2023Autonomous drilling of mouse cranial windows: evaluation on an egg modelIROS · Best Application Paper AwardEnduo Zhao (first author)Validation platform
2026Precision milling with general anatomical priors and active boundary sensingPreprint arXiv:2609.12530 · submitted to IEEE T-ROEnduo Zhao and Xiaofeng Lin (co-first authors), Yuhan SongRegistration 0.16 mm · live mouse 15/15

The first eight are peer-reviewed publications and the last row is a preprint; the manuscript on dynamic landmarks is not yet published and is not listed. The “paper topic” column summarizes each paper and is not its title. The team has 73 papers in this field, 15 of them core papers and 11 with a team member as first author; DOIs can be checked in the diligence pack.

Patent portfolio in detail
Patent certificate and application document 1 Patent certificate and application document 2 Patent certificate and application document 3 First page of the WIPO international publication

IP PORTFOLIO

The team holds 8 patents — 6 granted and 2 accepted for filing — across China, Japan, and WIPO

Counted by patent family (the Chinese and WIPO force-feedback filings count as one). The two patents that map directly onto the current product are the core navigation applications, both accepted for filing: the Japanese application for a navigation system, algorithm, and instruments (Japanese Patent Application No. 2025-284200) and the Chinese application for a 3D calibration board and performance-evaluation method for stereo microscopic imaging (CN202610805057.1). The six granted patents are in surgical robotics and reflect prior technical work; all eight name CEO Xiaofeng Lin as inventor. The barrier is structural: sub-millimeter intraoperative real-time registration × micron-scale real-time instrument tracking, combined in one hardware-software system. Ownership, licensing scope, and freedom to operate require item-by-item confirmation during investment and partnership due diligence.

Team

A serial-founder team at the intersection of medicine and engineering

R&D, algorithms and AI, clinical work, technology transfer, business and supply chain, company and capital — each core step in taking a device from the lab to the hospital has an owner. Core members have research and academic experience at the University of Tokyo, Tsinghua University, Harvard University, Yale University, the University of Chinese Academy of Sciences, and Harbin Institute of Technology; two advisors cover clinical translation and industry and registration.

Founder and CEO Dr. Xiaofeng Lin

FOUNDER & CEO

Dr. Xiaofeng Lin

Project Researcher in medical devices, Faculty of Medicine, the University of Tokyo · Leads product, technology, research and clinical work, and strategy

Experience

  • Project Researcher in medical devices, Faculty of Medicine, the University of Tokyo
  • PhD in biomedical engineering, the University of Tokyo
  • Master's degree, Shenzhen Institutes of Advanced Technology, University of Chinese Academy of Sciences
  • Former early co-founder of Hengle Medical Technology, a Series B+ company

Achievements

  • Global Medical Robotics Innovation Design Competition: Silver Award
  • 6th China International Internet+ College Student Innovation and Entrepreneurship Competition: Gold Award
  • IROS Workshop Best Poster Award in surgical robotics
  • Multiple top-venue papers on surgical robotics and surgical navigation
  • Eight patents in surgical robotics and surgical navigation
  • Guanhao Lin COO
  • Enduo Zhao CTO
  • Yuhan Song R&D Director
  • Zenghui Yu Business Director
  • Shenghao Jiang Clinical Research
  • Gina Quan Tech Transfer
  • Jian Lu Clinical Advisor
  • Chusheng Wu Industry Advisor
COO and co-founder Guanhao Lin

Guanhao Lin

COO · Co-founder

  • Former project manager in the strategy division of a multinational AI company; led Chinese-, English-, and Japanese-speaking teams building several AI products and led development of an enterprise AI-native system
  • Master's degree in systems innovation, the University of Tokyo
  • Bachelor's degree, School of Economics and Management, Harbin Institute of Technology
  • Former consulting associate at a venture capital firm
  • Role: company operations, fundraising, business, and regulatory registration
CTO and co-founder Enduo Zhao

Enduo Zhao

CTO · Co-founder

  • Postdoctoral researcher in biomedical engineering, Tsinghua University
  • PhD in mechanical engineering, the University of Tokyo; bachelor's degree, Department of Mechanical Engineering, Tsinghua University
  • Seven journal and conference papers on surgical robotics
  • IROS 2023 Best Application Paper Award (first author)
  • Role: R&D delivery, product engineering, and the R&D team
R&D Director and co-founder Yuhan Song

Yuhan Song

R&D Director · Co-founder

  • PhD candidate in biomedical engineering, the University of Tokyo; JSPS Research Fellow
  • Former Project Researcher at a Japanese national university
  • Four journal and conference papers in medical image processing
Business Director and co-founder Zenghui Yu

Zenghui Yu

Business Director · Co-founder

  • Former product manager on a Japan AI programme at a leading Chinese internet company
  • Former overseas business development manager at a Japanese manufacturer
  • Master's degree in commerce, Komazawa University
  • Several years as a manufacturing investment liaison at an investment institution
Clinical researcher Shenghao Jiang

Shenghao Jiang

Clinical Researcher

  • MD candidate, UNSW Sydney
  • Master's degree in computational science and engineering, Harvard University
  • Former vision-algorithm researcher at an AI autonomous-driving company
  • Eight journal and conference papers in medicine and robotics
Technology-transfer researcher Gina Quan

Gina Quan

Technology Transfer Researcher

  • Former hardware and software engineer at a US medical robotics company
  • Master's degree in biomedical engineering, Yale University
  • PhD candidate, Faculty of Medicine, the University of Tokyo
Clinical Translation Advisor Jian Lu

CLINICAL TRANSLATION ADVISOR

Jian Lu

Supports: the technology-transfer pathway; access to resources at Southeast University and Zhongda Hospital; clinical application and extension across departments

Appointments

  • Chang Jiang Scholar Distinguished Professor
  • Vice President, Zhongda Hospital, Southeast University
  • Young Chief Professor and PhD supervisor, Southeast University
  • Former General Manager of Hengle Medical Technology, a Series B+ company
  • Vice Chair, Interstitial Tumor Implantation Committee, Chinese Anti-Cancer Association
  • Member, Youth Committee of the Interventional Physicians Branch, Chinese Medical Doctor Association

Achievements

  • Principal investigator for ten provincial- and ministerial-level projects, including the Jiangsu Distinguished Young Scholars Fund, a National Key R&D Program subproject, and National Natural Science Foundation grants
  • More than 50 SCI papers as first or corresponding author, including work in Lancet Gastroenterology & Hepatology (cover), Nature Communications, and Journal of Hepatology
  • Sixteen granted or pending patents
  • First Prize, 2022 Jiangsu Science and Technology Award
  • Young Investigator awards from leading interventional societies in North America, Europe, and Asia-Pacific
  • ASCO GI Conquer Cancer Merit Award
  • Outstanding Young Scholar, Interventional Physicians Branch of the Chinese Medical Doctor Association
  • National Distinguished Physician: Emerging Young Talent
Industry Advisor Chusheng Wu

INDUSTRY ADVISOR

Chusheng Wu

Supports: access to industry and supply-chain resources; guidance on the quality system and registration preparation; industry strategy

Appointments

  • Vice President, China Association for Medical Devices Industry
  • President, Guangdong Medical Devices Industry Association
  • Executive President, Guangdong Association for the Promotion of Medical Science, Technology and Industry Integration
  • Former Chairman, Guangdong Bencao Pharmaceutical Group

Market · Competition · Funding & Milestones

Targeting the intersection of microscopic coverage, instrument tracking, and continuous intraoperative registration, entering the market through modules and software

The project remains in R&D and validation. One system runs in two modes: before approval, revenue starts through the Training & Assessment mode, OEM modules, and software licences, so commercialization need not wait for full-system approval; after approval the same system switches to Surgical Navigation mode and the business moves to proprietary systems, software, and consumables.

  1. 01PositionHow it differs from existing products: microscopic coverage, instrument tracking, and continuous registration, all three at once
  2. 02MarketOf 2.25 M procedures a year, the first cut is 512 K brain-tumor resections
  3. 03RevenueTraining, modules, and software licences bring revenue before approval
  4. 04Risks and milestonesThree risks, each with a matching action; seed commitments in place, 18 months to Pre-A

01 · Position

Navigation has to keep answering two questions throughout resection: where is the instrument and what tissue is this. Existing products each have their strengths: some already bring the navigation image into the microscope, and some achieve continuous registration on exposed bone. What this product does differently is to do three things at once — cover the field under the microscope, track the instrument itself inside the microscopic view, and keep registering during surgery as tissue deforms.

Product / country①Extends to microsurgery②Instrument tracking in the microscopic view③Continuous intraoperative registrationHow it registers
Proprio ParadigmUnited States No No Yes Continuous registration on rigid anatomyA camera array matches exposed bone to the CT model in real time, registering each vertebra rigidly; designed for open spine surgery.
Brainlab Microscope NavigationGermany · on Zeiss / Leica microscopes Yes No No Tracks the pose of the microscopeTracks microscope pose and overlays preoperative images in the eyepiece; the overlay comes from the single preoperative registration.
Medtronic StealthStation S8United States Yes No No One registration, updated by intraoperative rescanInfrared / electromagnetic tracking of a reference frame, with the navigation image injectable into the microscope eyepiece; registration is updated by rescanning during surgery.
Beyeonics OneIsrael · digital microscope N/A No No A digital microscope, positioned as visualizationThe surgical field is digitized and sent to a head-mounted display; it includes no image registration or navigation function.
7D FLASHUnited States No No No Registration in seconds, one snapshot at a timeA reference frame clamps the vertebra and visible light matches exposed bone; after the anatomy changes, a new snapshot updates it.
Huake PrecisionChina · neurosurgical robotics No No No One registration before the craniotomyStructured light scans the surface of the head and face to register; registration is done before the craniotomy and carried through after it.
Dynway intelligent navigation for microsurgeryThis product · meets all three within this table Yes Yes Yes Continuous registration that updates with tissue deformationRegisters to the tissue itself inside the microscopic view; no reference frame and no intraoperative rescan. Overall system accuracy and clinical value are still being validated.

Criteria: ① takes conventional (non-microscopic) surgical navigation as the baseline and asks whether the product goes on to cover the field under the microscope; Beyeonics One has no navigation function, so it is marked not applicable. ② means image-level tracking of the instrument itself inside the microscopic view, excluding tracking of external optical markers. ③ means registration is updated continuously and automatically during surgery from the anatomy itself, without rescanning; tracking only a rigid reference frame does not count. This product's intended specialties are neurosurgery, ENT, ophthalmology, and reconstructive surgery.

Compiled in November 2026 from each manufacturer’s public materials and registration information. Competitor specifications and approval status change over time; current specifications, indications, and registration status are subject to each manufacturer’s own disclosures. Product names and trademarks belong to their respective owners and are used here for identification only.

02 · Market → 03 · Revenue

  1. ADDRESSABLE 2.25Mprocedures/yr Core microsurgical procedures in China

    Neurovascular, malignant brain tumor related, ENT, and non-cataract ophthalmic procedures combined.

  2. LAUNCH INDICATION 512Kprocedures/yr Launch wedge: brain-tumor resection

    Boundary judgment directly determines the extent of resection and survival benefit, the clearest value proposition.

  3. FIRST SITES Leading tertiary Neurosurgical centers install first, after approval

    A target pool of about 120 leading neurosurgical centers; then ENT, ophthalmology, and reconstructive surgery.

  4. STAGE 0 · FROM 2027 Training mode Microsurgical skill training and assessment before approval

    Same hardware, with software enabled by mode. Education and research only, not for diagnostic or treatment decisions; for microsurgical skills centers at medical schools and teaching hospitals, microsurgery training centers, and animal laboratories. Training customers become the first installation candidates.

  5. STAGE 01 · 2028–29 Modules + licences Revenue during research / OEM validation

    Hardware module delivery; software licences bundled with hardware and billed annually; engineering services for microscope OEMs only, one-off per model. Customers start with university and research-institute laboratories, then microscope manufacturers and navigation-system vendors (including surgical-robot makers), validating demand, interfaces, and willingness to pay. Tertiary neurosurgery centers are the installation target after approval.

  6. STAGE 02 · POST-APPROVAL Systems + consumables Own brand at scale

    The same system switches to Surgical Navigation mode. Complete-system installations build scale; enhanced software licences renew annually; dedicated consumables launch after approval, forming recurring revenue.

TRAINING & ASSESSMENT MODE · FROM 2027

Skill training and assessment: turning experience into measurable indicators

  • Use: education and research, not for diagnostic or treatment decisions
  • Subjects: models, animals, and postoperative video, deliverable before approval
  • Output: instrument trajectory and steadiness, boundary recognition, and time; registration and trajectory data are passed to a large language model as interpretable context to generate targeted feedback and debriefs (the input is geometry, not black-box image scoring)
  • Training-stage data involve no patients, so they can be collected at scale first and fed back into the navigation algorithms
SURGICAL NAVIGATION MODE · ENABLED AFTER APPROVAL

Real-time intraoperative microsurgical navigation: boundary, instrument, and distance inside the existing view

  • Use: clinical; NMPA Class III, targeting submission acceptance in Q2 2029 and approval in 2030
  • Subject: the live surgical field, with dynamic registration following soft-tissue deformation
  • Target customers: leading tertiary neurosurgery centers, then ENT, ophthalmology, and reconstructive surgery
  • Same hardware and registration approach as the Training & Assessment mode, delivered as a registered product after approval

One set of hardware and one registration and tracking stack, with software enabled and charged by mode. Before approval, experience is turned into measurable indicators and revenue; after approval, the system enters the operating room. Commercialization and registration run in parallel, releasing risk in two steps.

Market breakdown: from the overall market to the launch wedge
Overall: the surgical navigation marketGlobal: USD 8.38 billion in 2024, forecast USD 18.47 billion in 2030, a 14.0% CAGR. China: USD 250 million in 2024, forecast USD 590 million in 2030 (about RMB 1.7 billion to 4.1 billion at 1:7), a 15.3% CAGR. China was about 3.0% of the global market in 2024, so penetration is low. Scope: all navigation systems and consumables; the drivers are minimally invasive, digital, and precision surgery raising demand for navigation.
The relevant part: microsurgeryAbout 2.25 million core microsurgical procedures a year in China: neurovascular about 1.154 million, brain tumor about 512 thousand, ENT about 301 thousand, and non-cataract ophthalmic about 284 thousand. By hospital pool: about 1,800 tertiary Grade-A hospitals, of which some 1,100 (60%) are assumed capable of microsurgery.
Launch: brain-tumor resectionThe sharpest pain point and the most consistent value proposition: boundary judgment directly determines extent of resection and survival benefit (+29.4 months for gross total resection). Leading tertiary neurosurgery centers come first, a target pool of about 120; ENT, ophthalmology, and reconstructive surgery follow. The entry channel is existing partner institutions — Sun Yat-sen University Cancer Center (Neurosurgery) and the University of Tokyo Faculty of Medicine — where installation partnerships are under discussion.

Sources: procedure volumes from the China Health Statistics Yearbook; market size from a published market report by Grand View Research. Third-party forecasts rest on their own definitions and assumptions and may differ from actual outcomes.

Beyond the first product: what else the same perception layer can do
Surgical skill training and assessment · from 2027Already on the revenue chain as STAGE 0: the same hardware with software enabled by mode, for medical-school skills centers, animal laboratories, and training centers.
More microsurgical specialtiesAfter neurosurgery come ENT, ophthalmology, and reconstructive surgery, as in FIRST SITES above; the same beam-split imaging and dynamic registration, with indications extended specialty by specialty as clinical validation progresses.
Automating brain-computer-interface implantation · candidate second product lineInvasive BCI implantation begins with a craniotomy and positioning step. The team’s autonomous mouse-craniotomy drilling system (IROS 2023 Best Application Paper Award) performs exactly that step, and micro-scale registration with fiber tracking is the prerequisite for automating it. We build the surgical layer only, not electrodes or decoders; BCI companies are customers of that layer. No resources are allocated to this line at present.
Perception layer for embodied surgical intelligence · long termThe two questions navigation answers, where the instrument is and what tissue this is, are the same two premises a surgical robot must establish before any action. The perception layer serves the surgeon first, then the robot.

04 · Risks, Funding and Milestones

The next priority is converting technical performance into clinical evidence and a registrable product. The innovation is concentrated in perception and registration; the system advises and never acts, and the decision always rests with the surgeon. Each of the three risks has a matching action, and exposure is released in stages.

RISK 01 · TECHNOLOGYThe core algorithm is validated in vivoInitial registration error of 0.16 mm with 15/15 live-mouse craniotomies; fiber tracking point measured at 9 µm on the bench. 74 animal studies to date; ex vivo results published in IEEE T-ASE (2025) and RA-L (2026), and the in vivo results submitted to IEEE T-RO. In vivo validation of soft-tissue deformation is not yet complete.Next: ex vivo porcine tissue to test adverse conditions such as bleeding and glare, then 6–12 in vivo porcine studies within 2027, forming the animal-study evidence for the registration submission; two integrated prototypes go to registration testing in 2027.
RISK 02 · APPLICATIONAdvises without acting, and degrades safelyThe system advises and never acts. Registration is non-invasive and needs no bony markers. On failure it degrades in three steps by registration confidence — reject points and hold, freeze and prompt, withdraw the overlay and hand back to the surgeon — with no silent correction. Clinical use adds a single registration of under five minutes and leaves the operating-room layout unchanged.Next: from 2027 the Training & Assessment mode is used on models and animals, with no patient contact; from 2028 patient contact occurs only within an ethics-approved small-sample clinical trial, releasing application risk in stages.
RISK 03 · REGISTRATIONThe category has precedent; the new method is evidenced on its ownSurgical navigation as a category already has approved precedents, and the testing standards are defined (GB 9706.1 and others). Compared with products on the market, the differences are concentrated in perception and registration. The uncertainty lies in sample size, number of centers, and review deficiency rounds.Next: perception and registration are evidenced with the product's own clinical data rather than by comparison with an equivalent device, planned around a single-arm supplementary clinical trial with an accuracy endpoint at 2–3 centers; pre-submission consultation with the review authority in Q4 2027 fixes the pathway and the Innovative Medical Device Special Review procedure is evaluated; deficiency-response buffer is already built into the 2030 approval target.

Seed-round investment commitments are in place from several individual investors, funding 18 months of work and carrying the company to Pre-A; prizes and government grants serve as non-dilutive supplements. The path runs from closing to prototype delivery, third-party validation, hospital proof-of-concept, and the start of Pre-A, with each deliverable tied to a disbursement point.

M0–3
Closing and start of operations

Counted from the closing date; IP work begins and the Guangzhou base and operating foundation are set up.

Start
M3–6
Two integrated V1 prototypes delivered

Two units suitable for demonstration and third-party testing; the microscope interface specification is finalized.

Delivered M6
M6–12
Third-party validation and registration groundwork

Phantom accuracy, biocompatibility, and electromagnetic-compatibility testing; classification consultation.

From M9
M12–18
Hospital validation and expert network

Target: proof-of-concept agreements with at least two hospitals and interface-validation talks with one OEM; Pre-A begins.

From M12
Resources and deliverables for the next stage

RESOURCE PRIORITIES

Three resource groups must be secured in parallel

Clinical validationAnimal studies, close clinician collaboration, data collection, and staged reviews.
Product engineeringHardware and software optimization, systems integration, industrial design, and prototype iteration.
Operations and complianceIntellectual-property strategy, regulatory planning, quality systems, manufacturing, and supply-chain management.
Team and external supportThe founding team moves to full-time in stages from 2027, covering R&D, clinical work, technology transfer, and operations; external support includes third-party testing bodies, patent agents and FTO counsel, regulatory and classification consultants, and animal-study contractors.

MILESTONES

Five deliverables for this round

Intellectual property and patent strategyPatent drafting and prosecution, FTO searches, and entry into China, Japan, and PCT.
Two integrated V1 prototypesHardware and software integration and commissioning, producing units suitable for demonstration and third-party testing, with the microscope interface specification finalized.
Third-party testing and registration groundworkPhantom accuracy, electromagnetic compatibility, and biocompatibility evaluation, plus classification consultation.
Hospital validation and expert networkJoint validation with clinicians, data collection and review, hospital proof-of-concept agreements, and interface discussions with microscope manufacturers.
Guangzhou base and operating foundationCore hires, quality-system development, and manufacturing readiness.

The basis and sources for procedure volumes are given in the Clinical Problem section above. Market size and commercial path illustrate the commercial logic; actual performance will depend on installation ramp, gross margin, channel discounts, and the post-registration launch schedule. Milestones are planning figures; actual dates depend on the closing date and validation progress. Specific pricing, installation plans, the round size, and detailed use of funds are calculated against the milestone budget, cash flow, and registration plan, and are not disclosed on this page. They are available in formal due-diligence materials.

CONTACT

Clinical, industry, and investment enquiries

We welcome direct contact from neurosurgical and microsurgical teams, microscope and instrument manufacturers, and investors following early-stage medical device programs.

Full due-diligence materials — covering the clinical validation plan, registration pathway detail, financial projections, and use of funds — are available on request. Please state your organization and intended use.

Enable JavaScript to view the email address
Dynway WeChat contact QR code WeChat · scan or long-press the code

DYNWAY

From microsurgical navigation to the perception layer of embodied surgical intelligence

Dynway's first product is an intelligent navigation system for microsurgery, built on two key innovations: sub-millimeter intraoperative real-time registration and micron-scale real-time instrument tracking. Together they form a layer of intraoperative perception — the foundation of the perception layer for embodied surgical intelligence, and what makes automated brain-computer-interface implantation possible. It serves microsurgical navigation first, then extends to surgical skill training and assessment.