Key Takeaways
- Supplying the emitting coil with a quadratic excitation voltage instead of a sinusoidal voltage reduced the maximal position error from 10.9 mm to 3.8 mm in the presence of a conductive distorter (aluminum disk).
- The root mean square error (RMSE) near the conductive distorter fell from 3.5 mm (sinusoidal) to 1.6 mm (quadratic) — a reduction of more than 50%.
- The method is adaptive to the amount of distortion and requires no analytical modeling of the conductive distorter's parameters, unlike prior lead–lag or RL-circuit compensation schemes.
- Quadratic excitation exploits the fact that the directly induced voltage depends on the first derivative of the emitting-coil voltage, while the distorter-induced voltage depends on the second derivative — allowing the two to be mathematically separated.
Clinical Context
Electromagnetic tracking systems (EMTS) are widely used in computer-assisted surgery — including ear, nose, and throat (ENT) navigation, cranial and spine procedures, cardiac mapping, bronchoscopy, and laparoscopic ultrasonography — because they require no line-of-sight and no rigid mechanical linkage. However, the operating room is a hostile electromagnetic environment: C-arms, LCD displays, operating microscopes, metal-studded walls, surgical instruments, and electronic devices all distort the reference magnetic field.
Ferromagnetic materials alter field shape and distort both DC and AC tracking, while conductive materials generate eddy currents that produce secondary AC magnetic fields. These distortions are difficult to characterize analytically and represent one of the most serious obstacles to clinical accuracy. Static distortions from fixed equipment can be addressed through calibration, but dynamic distortions from movable tools and imaging equipment cannot — and filtering or averaging introduces latency incompatible with real-time navigation.
This paper proposes a fundamentally different strategy: rather than modeling or filtering the distorter, it modifies the excitation waveform so that the undistorted signal can be isolated from the distorted one at the level of the sensing coil.
Clinical Evidence & Cohort Breakdown
The experimental validation used an 8-coil emitter (40 mm coil diameter), an NDI field sensor (0.3 mm diameter) as the sensing coil, and an aluminum disk (radius 0.05 m, thickness 0.03 m) as a worst-case conductive distorter — representing surgical tools such as hammers, trocars, endoscopes, or ultrasonic probes. The sensing coil was moved by a non-conductive robotic arm to 200 random positions over a 0.1 m × 0.1 m working surface at 0.15 m height. An optical tracking system (Polaris Spectra, RMSE < 0.25 mm) served as ground truth, with a 0.7 mm registration error between coordinate systems.
| Scenario | Excitation | Conductive Distorter | Max Error (mm) | RMSE (mm) |
|---|---|---|---|---|
| A | Quadratic | Absent | 3.4 | 1.7 |
| B | Sinusoidal | Absent | 1.5 | 0.9 |
| C | Quadratic | Present (Al disk) | 3.8 | 2.0 |
| D | Sinusoidal | Present (Al disk) | 10.9 | 2.4 |
Key observations:
- Under sinusoidal excitation, introducing the aluminum disk increased maximal error from 1.5 mm to 10.9 mm and RMSE from 0.9 mm to 2.4 mm — a severe degradation.
- Under quadratic excitation, the same distorter changed maximal error only from 3.4 mm to 3.8 mm and RMSE from 1.7 mm to 2.0 mm — an increase smaller than the measurement standard deviation.
- Error measurements under quadratic excitation showed no distance-dependent trend as the sensor moved away from the disk, whereas sinusoidal excitation showed a clear declining trend — confirming residual distortion sensitivity.
Intraoperative IONM Protocol & Technical Pearls
The following steps summarize the quadratic-excitation workflow as implemented in the experimental tracking system, framed for translation into an intraoperative navigation context.
- Model the coupled-coil system. Treat the emitting and sensing coils as magnetically coupled RL circuits. The directly induced sensing voltage is proportional to the first derivative of the emitting-coil voltage (transmittance
G₁₋₂(s) = M′₁₂·s/(1/τ₁ + s)). - Model the conductive distorter as an additional RL circuit. The indirectly induced voltage (via the distorter) is proportional to the second derivative of the emitting-coil voltage, yielding a high-pass characteristic (
G₁₋D₋₂(s)with time constantτ_D = L_D/R_D). - Apply quadratic excitation to the emitting coil (
U_Q(t) = U·t²). The directly induced signal becomes a sawtooth + rectangular pulse + exponential transient, while the distorter-induced signal becomes a rectangular pulse + exponential transient only — the missing sawtooth is the discriminator. - Estimate and subtract the transient response. Immediately after the quadratic pulse, feed a rectangular pulse into the emitting coil. Double-integrate the measured rectangular response to reconstruct the transient component of the quadratic response, then subtract it from the measured sensing-coil voltage.
- Extract the sawtooth gradient. Fit the residual sawtooth curve using least-squares; the gradient yields the mutual inductance
M₁₂between emitting and sensing coils. - Solve position and orientation. Use the recovered mutual inductance (free of distorter influence) with the known coil dimensions to compute sensor position and orientation via the standard position/orientation algorithm.
- Allow adequate acquisition time. The experimental protocol averaged 8 s of sinusoidal (1 kHz) data versus 32 s of quadratic data to achieve comparable signal-to-noise — a latency trade-off that must be weighed against real-time navigation requirements.
Technical pearls:
- The method is adaptive: no prior knowledge of the distorter's size, conductivity, form, or orientation is required.
- Aluminum was deliberately chosen over the less-conductive, non-ferromagnetic steel of real instruments to represent a worst-case distortion scenario.
- The approach is complementary to — not a replacement for — static calibration of fixed equipment and shielding strategies.
Clinical Implications for Surgeons & IONM Specialists
For teams performing image-guided ENT, cranial, spine, and bronchoscopic procedures, electromagnetic navigation accuracy is only as good as its immunity to the conductive instruments and equipment that surround the field generator. A tracking error that jumps from 1.5 mm to nearly 11 mm when a metallic tool enters the field is clinically unacceptable in skull base or sinus surgery, where critical neurovascular structures lie millimeters away.
Quadratic excitation offers a waveform-level solution that isolates the undistorted signal without requiring distorter modeling, dynamic filtering, or workflow-altering equipment repositioning. While the 32 s averaging window in this benchtop study is longer than desirable for live navigation, the underlying principle — separating first-derivative from second-derivative signal components — provides a robust foundation for next-generation distortion-immune trackers. Integration with IONM workflows, where continuous positional feedback must coexist with high-density electrophysiological recording, will require further reduction of acquisition latency before routine intraoperative deployment.
Reference
Bien T, Li M, Salah Z, Rose G. Electromagnetic tracking system with reduced distortion using quadratic excitation. International Journal of Computer Assisted Radiology and Surgery. 2014;9(2):323–332. DOI: 10.1007/s11548-013-0925-4.
Medical Disclaimer: This article is provided for educational and informational purposes only and does not constitute medical advice. Clinical protocols, device parameters, and surgical techniques described herein should be independently validated and adapted to individual institutional standards, regulatory requirements, and patient-specific considerations before clinical implementation.