💻 computer science

Automated Robotic Needle Puncture for Percutaneous Dilatational Tracheostomy

This paper presents a velocity-controlled robotic system guided by electromagnetic sensors and an adaptive constrained controller that significantly improves the accuracy and safety of automated needle punctures for percutaneous dilatational tracheostomy, reducing median position and angular errors to 1.7 mm and 4.13 degrees respectively in mannequin trials.

Original authors: Yuan Tang, Bruno V. Adorno, Brendan A. McGrath, Andrew Weightman

Published 2026-02-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yuan Tang, Bruno V. Adorno, Brendan A. McGrath, Andrew Weightman

Original paper licensed under CC BY 4.0 (http://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 you are trying to thread a needle through a tiny hole in a piece of fabric, but you are wearing thick gloves, the fabric is moving slightly, and you can only see the hole through a tiny, blurry peephole. That is essentially what doctors face when performing a Percutaneous Dilatational Tracheostomy (PDT).

In this procedure, a doctor needs to poke a needle through a patient's neck to create an airway for breathing support. It's a high-stakes game of "pin the tail on the donkey," where missing the target by a few millimeters or tilting the needle slightly can cause serious bleeding or damage to the windpipe.

Here is how this paper proposes to solve that problem using a robot, explained simply:

The Problem: The "Shaky Hand" and the "Blind Spot"

Currently, doctors do this manually. They feel for the right spot with their fingers and use a camera (bronchoscope) to peek inside the throat. But the camera only gives a 2D view of a 3D problem, like trying to park a car using only a side-view mirror.

The Result: Human hands shake, and depth perception is tricky. Doctors often miss the center by a wide margin (up to 5mm off-center or 30 degrees off-angle), which can be dangerous.

The Solution: A "Smart, Self-Correcting" Robot Arm

The researchers built a robotic system to take over the needle-poking job. Think of it as a super-precise, self-correcting drone that holds the needle.

Here is how the system works, broken down into three "magic tricks":

1. The "GPS" System (Electromagnetic Sensors)

Since the robot can't "see" inside the neck like a human, it uses two tiny magnetic sensors as its GPS:

Sensor A (The Target): Placed inside the patient's windpipe via a camera tube. It tells the robot exactly where the "bullseye" is.
Sensor B (The Needle): Attached to the tip of the needle. It tells the robot exactly where the needle is pointing.
The Magic: The robot constantly compares where the needle is versus where the bullseye is, adjusting its path in real-time.

2. The "Adaptive Learner" (Fixing Mistakes on the Fly)

Robots are usually rigid; if you move the robot's base slightly, it gets confused. But this robot is like a smart driver who learns the car's quirks instantly.

If the robot's base is slightly off from where it thinks it is, or if the needle is attached a tiny bit crookedly, the robot's "brain" (an adaptive controller) figures this out while it moves.
It constantly recalculates its own geometry, saying, "Oh, I thought I was here, but my sensors say I'm actually there. Let me adjust my steering." This happens so fast that the needle stays on course even if the setup isn't perfect.

3. The "Force Field" (Collision Avoidance)

The most important rule for a surgical robot is: Don't hit the patient.

The researchers programmed a virtual "force field" around the patient's neck.
Imagine the patient is inside an invisible bubble. If the robot arm tries to bump into the skin or the neck, the "force field" pushes it back gently.
It also creates a virtual "tunnel" (a cylinder) that the needle must stay inside. If the needle tries to wiggle out of the tunnel, the robot corrects it immediately. This ensures the needle goes straight in without stabbing the wrong side of the throat.

The Test Drive: 400 Practice Runs

To prove it works, the team didn't just simulate it on a computer; they tested it on a medical mannequin (a realistic dummy).

They set up the robot in different positions to mimic different patients.
They performed 400 needle punctures.
The Result: The robot was incredibly accurate.
- Position Error: It missed the center by only 1.7 mm on average (about the width of a pencil eraser).
- Angle Error: It tilted less than 4.5 degrees off-center.
- Safety: It never hit the "force field" or damaged the dummy.

Why This Matters

Currently, a human doctor needs to do this delicate task while another doctor holds the camera. This robot changes the dynamic:

One person can manage the airway (holding the camera).
The robot handles the scary part (poking the needle).
The Outcome: A safer, more consistent procedure that reduces the risk of bleeding and damage, especially for patients who are very sick and can't tolerate mistakes.

The Catch

The robot is only as good as the data it gets. If the person holding the camera tube inside the throat places the sensor slightly wrong, the robot will aim at that wrong spot. It's like a GPS guiding you to a destination, but if you typed in the wrong address, the GPS will still drive you perfectly to the wrong place.

In summary: This paper presents a robotic "autopilot" for a dangerous surgery step. It uses magnetic sensors to see, math to learn its own mistakes, and virtual force fields to stay safe, turning a shaky, high-risk human task into a precise, repeatable machine operation.

1. Problem Statement

Percutaneous Dilatational Tracheostomy (PDT) is a critical procedure for patients in intensive care units requiring prolonged mechanical ventilation. The most dangerous and critical step is the needle puncture, which must be performed accurately through the anterior tracheal wall between specific tracheal rings.

Current Challenges: Current manual practices rely on palpation and 2D visualization (bronchoscopy or ultrasound). These methods lack accurate depth perception and 3D navigation, leading to significant position errors ( $\ge$ 5 mm) and angular errors ( $\ge$ 30°).
Consequences: Inaccurate punctures can cause major bleeding, posterior tracheal wall perforation, thyroid damage, or tube misplacement.
Gap: Existing robotic solutions for PDT are rare and often utilize an "inside-out" approach. There is a need for an automated "outside-in" robotic system that ensures high precision, avoids collisions with the patient, and adapts to kinematic uncertainties.

2. Methodology

The authors propose a robotic system guided by electromagnetic (EM) sensors and controlled by an Adaptive Constrained Controller.

A. System Hardware

Robot: A 7-DoF Franka Emika serial manipulator.
Sensing: An electromagnetic tracking system with:
1. A transmitter near the patient's head.
2. A sensor attached to the needle tip for navigation.
3. A sensor attached to the bronchoscope tip inside the trachea to define the target position and direction.
Procedure: The robot performs three steps: (1) Aligning the needle with the guide line, (2) Moving to a safe start position above the neck, and (3) Advancing the needle to the target.

B. Control Framework: Adaptive Constrained Control

The core innovation is a control law based on Vector Field Inequalities (VFIs) combined with Adaptive Control.

Nominal Constrained Control (VFIs):
- Uses differential inequalities to define safe zones and collision avoidance constraints directly in the control input.
- Collision Avoidance: Defines a cylinder around the patient to prevent the robot from hitting the body.
- Target Constraints: Defines a guiding cylinder and cone around the target to ensure the needle stays within a safe angular and radial deviation.
- Posterior Wall Protection: Defines a plane constraint to prevent the needle from penetrating too deeply and damaging the posterior tracheal wall.
Adaptive Control Law:
- Addresses uncertainties in the system, such as the exact placement of the EM transmitter relative to the robot base and the mechanical attachment of the needle to the end-effector.
- The controller estimates and adapts these kinematic parameters online using real-time sensor feedback, ensuring the needle tip converges to the measured target despite initial calibration errors.

3. Key Contributions

Novel Robotic System: Development of the first automated "outside-in" robotic system for PDT needle puncture, replacing manual insertion.
Adaptive Constrained Controller: Integration of VFIs for safety (collision avoidance) with an adaptive law to handle uncertain kinematic parameters (robot-base offset and needle attachment) in real-time.
Comprehensive Validation:
- Simulation: 100 trials in CoppeliaSim validating the controller's ability to handle noise and parameter uncertainty.
- Experimental: 400 punctures performed on a mannequin with silicone skin and a tracheal model, simulating various patient anatomies.

4. Results

The system was evaluated against clinical requirements (Position error < 5 mm, Midline deviation < 30°).

Simulation Results:
- Mean position error: 1.55 ± 0.40 mm.
- Mean angular error: 0.38 ± 0.021°.
- The controller successfully adapted to large initial uncertainties (up to $\pm$ 200 mm base position error) and maintained constraints.
Experimental Results (400 Trials):
- Position Accuracy: The absolute median puncture position error was 1.7 mm (IQR: 1.9 mm).
- Angular Accuracy: The median midline deviation was 4.13° (IQR: 4.55°).
- Success Rate: 95.47% of punctures successfully passed through the target window.
- Constraint Adherence: The system maintained the needle within the guiding cylinder (radius 1.5 mm) and cone (angle 0.5°) with high precision.
- Correlation Analysis: Strong positive linear correlations were found between bronchoscope sensor measurement errors and final needle errors ( $r=0.909$ for position, $r=0.999$ for angle), indicating that the robot's accuracy is limited primarily by the manual placement of the bronchoscope sensor, not the robot's control.

5. Significance and Future Work

Clinical Impact: The system significantly reduces position and angular errors compared to manual techniques (which often exceed 5 mm and 30°). It automates the most dangerous step of PDT, potentially reducing complications like bleeding and tracheal wall damage.
Workflow Improvement: The system allows a bronchoscopist to manage the airway while the robot performs the puncture, reducing the need for a second operator to hold the needle.
Limitations: The accuracy is currently dependent on the manual skill of placing the bronchoscope sensor. The experiment used a mannequin with silicone skin, which differs from human tissue properties.
Future Directions:
- Developing mathematical models to calculate target positions rather than relying solely on manual sensor measurement.
- Automating the subsequent dilation step.
- Integrating haptic sensors to detect tissue resistance and prevent excessive force.
- Testing on animal tissues (porcine) for better simulation of human anatomy.

Conclusion: The study demonstrates the feasibility of using an adaptive constrained robotic controller to perform automated PDT needle punctures with high precision and safety, offering a promising solution to improve patient outcomes in critical care settings.