PRE-CLINICAL INVESTIGATIONAL PLATFORM Bench & retrospective surgical pilot validation (n=3 datasets, 2,654 frames) · Not yet cleared by the US FDA or CE for human clinical use
SYS-SPEC // MEDTECH RESEARCH / PRE-CLINICAL MOTION COMPENSATION PLATFORM

The Autonomous Motion Navigation Platform for Interventional Fluoroscopy.

An intelligent motion-compensation co-pilot that locks onto surgical targets in real time and automates radiation shielding during complex ERCP and interventional radiology.

Up to 78% Radiation Reduction (Modeled ALARA)
< 0.5 mm Mean Tracking Error (Pilot Dataset)
< 14 ms Unified Multi-Task Latency
0 Resets Manual Interruption Tax in Feeds
Omni Rovis Surgical Motion Compensation Workstation [Bench Verification Cockpit] ● BENCH VERIFIED (PILOT DATA)
Omni Rovis Desktop Surgical Cockpit Interface
01 / CLINICAL & DOSIMETRY UNMET NEED

The Hidden Stakes of Prolonged Fluoroscopy

Complex Endoscopic Retrograde Cholangiopancreatography (ERCP)—such as difficult bile duct stones (>15 mm) or stricture cannulation—imposes severe physical, radiation, and cognitive burdens on surgical teams.

CR-01 // DOSIMETRY

Radiation Exposure Burden

Published literature indicates complex therapeutic ERCP procedures often reach ~80+ Gy·cm² in Dose Area Product (DAP). Cumulative scatter elevates risks for cataracts and malignancies, while 15 kg lead aprons contribute to chronic surgeon orthopedic spine disease.

~82.4 Gy·cm² Published Conventional Mean
CR-02 // KINEMATICS

The "Manual Reset Tax"

Under sedation, natural diaphragmatic excursions (14 BPM) and involuntary patient coughs displace the anatomical region of interest (ROI) by up to 79 mm. The surgeon must pause, drop instruments, and manually re-jog the C-arm or table, compounding fatigue.

12–18 Pauses / Complex Case
CR-03 // OCCLUSION

Contrast Washout Blindness

Radiopaque iodine dye dissipates rapidly under continuous physiological bile flow. Once contrast clears, standard 2D tracking algorithms lose target visibility, forcing repeated contrast boluses or operating with degraded spatial certainty.

< 4.5s Typical Peak Dye Visibility
PEER-REVIEWED MOTION EVIDENCE (AAPM TG-76 / INTERVENTIONAL LITERATURE) VERIFIED CLINICAL SOURCES
4.9 – 30.4 mm Liver & Biliary Target Drift (Tidal Respiration)
10 – 20 mm Diaphragmatic Excursion (14 BPM Baseline)
8.0 – 13.7 mm Standard Radiation Safety Margin Padding
02 / ROBOTIC & PERCEPTION ARCHITECTURE

How Omni Rovis Achieves Autonomous Navigation

A unified multi-task perception backbone combined with Levenberg-Marquardt damped Image-Based Visual Servoing (IBVS) and patient-specific Bayesian motion calibration.

Unified Multi-Task Vision Architecture (Shared Backbone)

A single TensorRT feature extractor driving 5 lightweight task heads simultaneously, eliminating redundant compute and achieving sub-14ms closed-loop latency.

TASK HEAD 1
Detector Boundary & Origin
Global Coordinate Calibrator

Locks onto global intensifier / Flat-Panel Detector (FPD) boundary to establish the physical image origin [270.0, 270.0].

99.2% Lock (Pilot Feeds)
TASK HEAD 2
Skeletal Vertebral Anchors
Pedicle Landmark Extractor

Tracks rigid vertebral pedicle shadows and diaphragm curvature to form the spatial foundation for zero-contrast correlation.

96.8% Lock (Pilot Feeds)
TASK HEAD 3
Duodenoscope Tip & Elevator
Endoscope Articulation Head

Binds the distal duodenoscope articulation point at the ampulla, providing continuous relative mechanical grounding.

98.1% Lock (Pilot Feeds)
TASK HEAD 4
Curvilinear Guidewire Spline
Topology Vector Extractor

Extracts sub-pixel spline vectors tracing 0.035" hydrophilic guidewires through tortuous biliary duct geometry.

92.4% Spline Lock (Pilot)
TASK HEAD 5
Biliary Target Stone & Basket
Oriented Bounding Box Head

Computes 2D centroid and angular orientation theta of Dormia lithotripsy baskets, biliary stones, and stricture bifurcations.

94.7% Spatial Lock (Pilot)
ROBOTIC VISUAL SERVOING CONTROL LAW
v_c = -\lambda(e) \cdot \mathbf{L}_s^{+} \cdot (\mathbf{s} - \mathbf{s}^*) \quad \text{where} \quad \mathbf{L}_s^{+} = (\mathbf{L}_s^T \mathbf{L}_s + \mu \mathbf{I})^{-1} \mathbf{L}_s^T

Levenberg-Marquardt Damped Pseudoinverse: Calculates optimal instantaneous 3D velocity commands (\(v_c\)) for the motorized TruBlock collimator and C-arm. An adaptive gain law \(\lambda(e) = \lambda_{\text{base}} \cdot [0.80 + 0.20 \tanh(e/8.0)]\) enables rapid response to gross spasms (\(85\text{ mm/s}\)) while ensuring sub-millimeter damping near center without mechanical overshoot.

SYS-01 // BAYESIAN FUSION

Pulse-Recalibrated Motion Fusion

When contrast clears, a Bayesian Kalman Filter updates the target estimate using a patient-specific motion correlation matrix anchored to the L2 spine. The model recalibrates on every contrast pulse, maintaining an explicit confidence envelope.

SYS-02 // TRUBLOCK RETROFIT

TruBlock Standalone Collimator

Designed as a standalone beam-limiting add-on module requiring zero proprietary gantry write-access. Motorized lead shutters actively track the target, providing up to ~78% ALARA radiation shielding across peripheral tissue.

SYS-03 // DUAL INTERLOCK

Dual-Interlock Safety Architecture

Redundant safety gating: (1) Hardware Dead-Man Foot Pedal clamps mechanical brakes in <12 ms; (2) AI Confidence Floor automatically freezes actuation if multi-head confidence drops below 70%, surfacing an immediate lost-lock alert.

03 / RETROSPECTIVE PILOT EVIDENCE

Retrospective Pilot Studies & Bench Benchmarks

Evaluated across three retrospective surgical video datasets comprising 2,654 ground-truth annotated frames.

PILOT BENCHMARK #01 646 FRAMES @ 30 FPS · 0.24mm PILOT ERROR

Lithotripsy & Biliary Stone Extraction (Retrospective)

Retrospective video evaluation of single-operator cholangioscopy mechanical lithotripsy for an impacted 16.5 mm bile duct stone. Under simulated 14 BPM respiration, the Dormia extraction basket oscillates across a 42 mm sweep.

0.24 mm Mean Steady Tracking Error
100% Basket Lock in Pilot Clip
~79% Modeled Area DAP Shielding
  • Endoscope Head: Continuous lock on duodenoscope elevator in the duodenal lumen.
  • Oriented Box Head: Exact 2D bounding of Dormia basket during wire tensioning.
  • Guidewire Spline: Traces guidewire ascending through the papilla into the intrahepatic duct.
Clinical Clip 1 Lithotripsy Tracking Pilot Clip 1: Lithotripsy basket and guidewire locked under simulated closed-loop IBVS.
PILOT COMPARISON #02 292 FRAMES @ 30 FPS · STATIC VS DYNAMIC

Benchmark: Static ROI Shuttering vs. Omni Rovis

Comparative evaluation against standard static ROI collimator behavior (fixed box baseline) during selective cannulation of the major duodenal papilla.

Dynamic Follow Omni Rovis Active Tracking
Fixed Box Static Shutter Baseline
+28.5% Modeled Protection Gain
  • Static Baseline Limitation: Fixed shutter requires manual operator repositioning whenever anatomy drifts.
  • Omni Rovis Advantage: Autonomous dynamic collimator shutters actively follow the catheter, maintaining protective coverage.
Clinical Clip 2 Benchmark Pilot Clip 2: Omni Rovis dynamic collimation simulation vs. static background shutter.
PILOT BENCHMARK #03 1,716 FRAMES @ 30 FPS · 71.5mm STRICTURE

Complex Biliary Stricture Navigation (71.5 mm)

Surgical suite workflow constraints during complex 71.52 mm proximal Common Bile Duct stricture dilation. Retrospective evaluation under severe noise and tool overlap.

71.5 mm Stricture Path Traversed
95.0% Guidewire Spline Confidence
0 Loss Trajectory Lock Disruption
  • Bifurcation Lock: Precision lock on primary hepatic duct bifurcation stricture.
  • L2 Spine Anchor: Maintains spatial orientation during simulated table tilts and respiration.
Clinical Clip 3 Surgical Suite Constraints Pilot Clip 3: 71.5mm CBD stricture traversal with multi-task perception backbone.
04 / MODELED CLINICAL & ECONOMIC IMPACT

Dosimetry Modeling & Surgical Suite ROI

Modeled radiation dose reductions (ALARA Tier 1) and operational time savings projected across standard high-volume interventional suites.

CLINICAL RADIATION DOSE COMPARISON UP TO 78% SCATTER REDUCTION
Conventional Fluoroscopy vs TruBlock Active Collimation Radiation Dose Comparison

Conventional Fluoroscopy (Uncollimated Flood)

Wide conical beam floods the patient's entire abdomen, scattering heavily across the room and exposing the surgeon and nursing staff to high cumulative radiation.

TruBlock Active Collimation (Targeted Protection)

Autonomous motorized lead shutters dynamically narrow the beam cone onto the biliary target, shielding peripheral organs and cutting occupational scatter by up to ~78%.

Performance Metric Conventional Fluoroscopy Static ROI Shuttering Baseline Omni Rovis (TruBlock Mode)
Mean DAP Exposure / Case ~82.4 Gy·cm² (Literature) ~42.1 Gy·cm² ~17.8 Gy·cm² (Up to -78%)*
Occupational Eye Scatter (Staff) ~528 µGy / case ~295 µGy / case ~116 µGy / case (-78%)*
Patient Motion Follow None (Manual Jog) Manual Box Drag Autonomous Follow (<14ms latency)
Contrast Washout Handling Zero (Visual Loss) None Bayesian Spine Motion Fusion
Manual Reset Interruption Rate 12–18 pauses / case 8–12 pauses / case 0 pauses (Eliminated in Pilot)
Hardware Actuation Safety Manual Foot Switch Touchscreen UI Dual: Dead-Man Pedal + Confidence Freeze
* Modeled radiation shielding projection based on TruBlock dynamic lead aperture restriction over active surgical ROI. Baseline conventional ERCP DAP from published multicenter dosimetry studies (PubMed 22050826 ↗).
~$142,000
Projected Annual OR Savings

Modeled across 450 ERCP cases/year via estimated 8.5 minute reduction in manual C-arm re-centering and fluoroscopy dead-time.

Ergonomic
Interventionalist Relief

Designed to reduce reliance on heavy lead shielding during prolonged procedures, supporting surgeon orthopedic health.

ISO 13485
Regulatory Framework Target

Architecture designed to comply with IEC 62304 Class C ↗ medical software life-cycle and ALARA radiation minimization guidelines.

05 / HARDWARE INTERFACE & GTM STRATEGY

Two-Phase Go-to-Market & Systems Architecture

A pragmatic commercial strategy starting with a standalone retrofit collimator, expanding into OEM gantry navigation partnerships.

THE WINNING HARDWARE PARADIGM

How TruBlock Works: Zero OEM Tampering, 100% Beam Steering

Instead of attempting to command a 300 kg hospital machine, TruBlock leaves the C-arm completely frozen and drives 4 high-speed lead shutter blades to steer the X-ray beam directly.

01
READ-ONLY INGESTION

Standard HDMI Video Feed

Taps into the C-arm monitor's auxiliary Video-Out / DVI port. 100% passive and read-only. Zero modifications to proprietary Siemens, GE, or Philips firmware.

02
REAL-TIME PERCEPTION

Sub-14ms Target Centroid

Unified multi-task TensorRT vision backbone tracks the duodenoscope, guidewire, and moving biliary stone, calculating the exact spatial drift vector.

03
MICRO-STEERING MOTION

Lead Shutter Beam Aiming

Four NEMA 14 stepper motors slide 2.5 mm lead blades across miniature linear rails. The 300 kg machine stays frozen; the beam follows the stone!

EXPLODED CAD ASSEMBLY VIEW 6-LAYER PRECISION STACK
TruBlock Exploded CAD View

Layer-by-Layer Electromechanical Stack

Quick-release universal mounting collar, brushed 6061-T6 aluminum housing, 4x NEMA 14 micro-steppers with T8 lead screws, 4x overlapping 2.5mm lead blades on MGN9 rails, and custom STM32 motion controller PCB.

OPERATING SUITE RETROFIT BOLT-ON TO ANY C-ARM
TruBlock Mounted on Hospital C-Arm

Universal C-Arm Compatibility

Clamps securely onto standard 80–140mm tube ports in under 60 seconds without tools. Compact 8 kg module footprint with single umbilical cable routing to the mobile workstation console.

TruBlock Internal Mechanism Cross Section
INTERNAL ELECTROMECHANICAL LAYOUT

Whisper-Quiet, High-Speed Lead Shutter Actuation

Suspended on dual stainless steel MGN9 linear guide carriages and driven by TMC2209 silent microstepping drivers. A 5 mm respiration stroke is traversed in <35 milliseconds with zero audible clicking in the sterile operating field.

<35 msPhysical Step Response
0.05 mmMicrostepping Accuracy
<8.5 kgTotal Module Weight

Phase 1: TruBlock Retrofit (Beachhead)

Phase 2: OEM Gantry Licensing

  • Robotic Motion: CANopen / J1939 telemetry interface for next-gen motorized C-arms.
  • OEM Partnerships: Licensing visual servoing stack to Siemens, GE, and Philips.
  • Full Suite Integration: Complete patient-following motorized gantry autopilot.

Real-Time Safety & Ingestion

  • Video Stream: 10-Bit DICOM / GigE Vision ingestion at 30/60 FPS.
  • Inference: Sub-14ms unified multi-task TensorRT execution.
  • Dual Interlock: <12ms mechanical brake clamp + AI uncertainty freeze.
INDUSTRY-STANDARD VALIDATION (AAPM TG-76 / TG-324 PROTOCOL)

5-Stage Validation & Commercialization Roadmap

Adopting the programmable dynamic respiratory motion phantom methodology established by leading robotic radiotherapy systems (e.g. Accuray CyberKnife) to prove sub-millimeter tracking accuracy against known encoder ground truth.

STAGE 1 ✓ COMPLETED

Multi-Task Vision on Retrospective Feeds

Trained and benchmarked unified multi-task TensorRT vision backbone across 2,654 ground-truth surgical frames (<14ms latency, 0.24mm pilot error).

View 3 Ground-Truth JSONs ↗
STAGE 2 ● IN PROGRESS

AAPM TG-76 Dynamic Motion Phantom Rig

Closed-loop visual servoing validation on a programmable motorized respiratory stage driven by recorded asymmetric human breathing traces.

AAPM TG-76 Protocol (PDF) ↗
STAGE 3 BEACHHEAD PRODUCT

TruBlock Standalone Collimator Prototype

Benchtop integration of autonomous motorized lead shutter retrofit module (Class II 510k Beam-Limiting Device regulatory pathway).

FDA 21 CFR 892.1610 ↗
STAGE 4 ACADEMIC BENCH

University Research C-Arm Partnership

In vitro phantom evaluation on academic interventional radiology research units with open API control access.

IEC 62304 Medical Standard ↗
STAGE 5 COMMERCIAL SCALE

OEM Gantry Autopilot Licensing

Commercial software licensing and CANopen integration with major global C-arm manufacturers (Siemens, GE, Philips).

OmniIDE Research Group ↗
06 / LEADERSHIP & CLINICAL COLLABORATION

Leadership & Institutional Partnerships

Omni Rovis is developed by the AI research and robotics division at OmniIDE, based in Bengaluru, India. We are actively engaging with interventional gastroenterologists, hospital research institutions, and MedTech investors.

Mohammed Nihan — Founder & CEO
EXECUTIVE LEADERSHIP

Mohammed Nihan

Founder & Chief Executive Officer

Founder & CEO of OmniIDE and Omni Rovis. Leading AI architecture, real-time visual servoing control systems, and algorithmic medical motion compensation infrastructure.

OmniIDE Company Logo
PARENT ORGANIZATION

OmniIDE Engineering & Research

Bengaluru, Karnataka, India · MSME Registered

AI research and engineering organization building autonomous intelligence systems and medical robotics software. Registered enterprise (UDYAM-KR-03-0681404).