# Fiber optical shape sensing of flexible instruments for endovascular navigation

- 170 Downloads

## Abstract

### Purpose

Endovascular aortic repair procedures are currently conducted with 2D fluoroscopy imaging. Tracking systems based on fiber Bragg gratings are an emerging technology for the navigation of minimally invasive instruments which can reduce the X-ray exposure and the used contrast agent. Shape sensing of flexible structures is challenging and includes many calculations steps which are prone to different errors. To reduce this errors, we present an optimized shape sensing model.

### Methods

We analyzed for every step of the shape sensing process, which errors can occur, how the error affects the shape and how it can be compensated or minimized. Experiments were done with one multicore fiber system with 38 cm sensing length, and the effects of different methods and parameters were analyzed. Furthermore, we compared 3D shape reconstructions with the segmented shape of the corresponding CT scans of the fiber to evaluate the accuracy of our optimized shape sensing model. Finally, we tested our model in a realistic endovascular scenario by using a 3D printed vessel system created from patient data.

### Results

Depending on the complexity of the shape, we reached an average error of 0.35–1.15 mm and maximal error of 0.75–7.53 mm over the whole 38 cm sensing length. In the endovascular scenario, we obtained an average and maximal error of 1.13 mm and 2.11 mm, respectively.

### Conclusion

The accuracies of the 3D shape sensing model are promising, and we plan to combine the shape sensing based on fiber Bragg gratings with the position and orientation of an electromagnetic tracking to obtain the located catheter shape.

## Keywords

Fiber Bragg grating (FBG) Shape sensing Flexible instruments Endovascular navigation## Introduction

Cardiovascular diseases are the main cause of death in western industrial nations [9]. Some of these diseases like abdominal aortic aneurysms can be treated with an endovascular aortic repair (EVAR) procedure, in which a stent graft is placed in the aneurysm region under 2D fluoroscopy. To reduce the X-ray exposure and to supersede the angiography, a three-dimensional navigation is needed.

Fiber Bragg grating (FBG)-based systems are used for shape sensing, which enables three-dimensional navigation. FBGs are interference filters inscribed into the core of an optical fiber, which reflect a specific wavelength. Combining multiple FBGs at the same longitudinal position allows to calculate curvature and direction angle. The most common configurations are three fibers arranged triangular around the structure to be measured [4, 13]. This introduces significant errors due to possible changes in the core geometry [4], which can be overcome with multicore fibers, where several cores are integrated into one fiber [10]. In addition, other FBG systems with other geometries have been introduced, for instance helically wrapped [17].

Most research groups use FBG systems for shape and force sensing of medical needles [11], which have a simple bending profile allowing only shapes with low bending and no torsion. For example, Park [12] applied FBGs for shape sensing of biopsy needles and Roesthuis used it to reconstruct the shape of a nitinol needle [13]. A few works using optical fibers for flexible instruments have been reported in the literature: Shi [15] used FBGs for catheter shape sensing. Also Khan [7] used four multicore fibers to reconstruct the first 118 mm of a catheter. However, to our knowledge, there are currently no studies on the accuracy of fiber optical shape sensing for very long and flexible medical instruments.

In general, shape reconstruction of flexible structures is more challenging, because higher deflections and torsion can occur. Thus, the error analysis of the shape reconstruction from measured wavelengths to the reconstructed shape becomes more important. Also, the shape accuracy has to be very accurate, since the error accumulates along the fiber.

## Material and methods

We consider a multicore fiber with *n* FBG arrays along a flexible instrument, as shown in Fig. 1. Each array contains seven FBGs, one center core and six outer cores. All FBGs have fixed length \(\ell \), and the arrays are uniformly distributed with center-to-center distance *d*.

### Shape sensing model

- 1.
Wavelength shift calculation.

- 2.
Strain computation for every core.

- 3.
Strain interpolation for every core.

- 4.
Curvature and angle calculation.

- 5.
Curvature and angle correction.

- 6.
Shape reconstruction with circle segments.

#### Wavelength shift calculation

#### Strain computation

#### Interpolation

When the curvatures and angles are calculated for every FBG array, the intermediate values can be determined by interpolation. Henken [4] compared common interpolation methods for shape sensing and concluded that cubic spline interpolation is the best solution, which is currently the state-of-the-art interpolation. Interpolating the curvature is straight forward since it is continuous for every shape, whereas the angle interpolation is challenging for flexible structures, which may have discontinuous angle. Thus, we suggest to interpolate the strain, since it is continuous.

Furthermore, it is assumed that the measurements of one FBG array are the values for one specific position, usually the array center. Thus, we use the averaged cubic interpolation, as introduced in [6]: this yields a realistic interpolation based on the spatial properties of a FBG by modeling the measured value as an average over the sensor range.

#### Curvature and angle computation

The calculation of the curvature and direction angle depends on the fiber system. The most common one is a triplet configuration [4, 13]: here, the FBG system has three fiber cores with specific angles (typically \(120 ^\circ \)) in between, as illustrated in Fig. 1.

*x*. By solving the equation system, we obtain the strain bias \(\varepsilon _0\), the curvature \(\kappa \) and the direction angle \(\varphi \). The equation system can also be extended for four or more fibers.

The equations show that the curvature is influenced by the radii \(r_x\) similarly as by the photo-elastic coefficient. In addition, the strain is also biased by a temperature change, additional axial strain and pressure. Due to the short distance (\(< 100~ \upmu \mathrm{m}\)) of the FBGs in one array, it can be assumed that for every grating in one array (see Fig. 1), this bias is equal and therefore compensated by the strain bias \(\varepsilon _0\).

#### Curvature and angle correction

*c*to get the right curvatures

#### Shape reconstruction

*T*the tangent vector,

*N*the normal vector and

*B*the binormal vector of the curve at length position

*t*. The integration of the determined tangent vectors yields the shape of the curve. This method fails at points with \(\kappa = 0\) since the torsion is undefined there. Thus, this algorithm is not suitable for shape sensing of flexible structures.

*T*. The shape reconstruction is conducted in the same way as with Frenet–Serret.

Roesthuis [13] proposed another method based on circle segments: the shape is reconstructed by approximating it with elements of constant curvature. For every element a circle segment of curvature \(\kappa \) and length *l* is created and is rotated by the direction angle \(\varphi \). By repeating this procedure for every given set \((\kappa , \varphi )\) we obtain the whole shape.

### Experimental methods

For all experiments described below we had one multicore fiber system (FBGS Technologies GmbH) available with 7 cores, one center core and six outer cores with an angle of \(60^{\circ }\) in between, as shown in Fig. 1. It has 38 FBG arrays each with 5 mm length and 10 mm center-to-center distance, which are chains of draw tower gratings (DTG\(\circledR \)).

In the next sections, we used the following parameters and algorithms if they are not analyzed or specified otherwise: we fixed our covered fiber to a precise ruler and used the measured wavelength as reference wavelengths, we used a photo-elastic coefficient \(p_\mathrm{e} = 0.22\), made averaged cubic strain interpolation, used four outer cores and reconstructed the shape with circle segments.

#### Wavelength shift computation

For our multicore fiber, we had no reference Bragg wavelengths given. Thus, we had to determine these wavelengths with a measurement without any strain. Therefore, we analyzed the effect of the Bragg wavelength estimation: at different times we fixed the fiber in a straight line, measured the wavelengths, used it as reference Bragg wavelengths and reconstructed various types of shapes.

#### Strain calculation

Results of the Bragg wavelength study: measured errors \(e_{\text {avg}}\) and \(e_{\text {max}}\) in mm for different shapes using various Bragg wavelengths

Shape | Error | First reference | Second reference | Third reference |
---|---|---|---|---|

Straight line | \(e_{\text {avg}}\) | 0.36 | 0.16 | 2.28 |

\(e_{\text {max}}\) | 1.00 | 0.30 | 5.49 | |

Bended curve | \(e_{\text {avg}}\) | 1.70 | 1.56 | 1.59 |

\(e_{\text {max}}\) | 4.92 | 4.53 | 4.82 | |

S-curve | \(e_{\text {avg}}\) | 1.80 | 1.76 | 1.38 |

\(e_{\text {max}}\) | 4.58 | 4.62 | 3.06 |

#### Interpolation

We formed our fiber to a snakelike shape, which has a few singularity points, interpolated the measured strains as proposed in “Interpolation” section and compared the resulting curvature and angles with the ones of common interpolation methods.

#### Curvature and angle computation

Since we have a multicore fiber with six outer cores and one center core and an interrogator, where we can connect four cores, we do not have to use a triplet configuration with \(120^{\circ }\) in between. Thus, we analyzed the effect of different core configurations on the resulting curvatures and angles.

#### Curvature and angle correction

To determine the twist angle \(\varphi _{\text {twist}}\), we bent our fiber to a 2D-shape, where every position has the same angle and used the determined angles as twist angles, as described in Eq. (3). To get the curvature scale factor *c* of our fiber, we made bow shapes with different radii, determined the best value assuming a photo-elastic coefficient \(p_\mathrm{e} = 0.22\) and used it for curvature correction, as described in Eq. (2).

#### Shape reconstruction

The shape reconstruction quality depends completely on the input: when the measured values are correct, the proposed algorithms can reconstructed the correct shape. Therefore, we analyzed the following two aspects:

First, we looked at the convergence, i. e. how fine the segments in each step must be for accurate shapes. Second, we analyzed the noise handling of the three algorithms, i. e. how the resulting shape change with increasing Gaussian noise. In both cases, we simulated an arc shape with torsion, calculated the average curvature and median angle for every segment and reconstructed the shape.

#### 3D shape reconstruction accuracy

To evaluate our model, we recorded 3D measurements: we covered our fiber (diameter: \(200\,\upmu \mathrm{m}\)) with a metallic capillary tube (inner diameter: \(300\,\upmu \mathrm{m}\), total diameter: \(400\,\upmu \mathrm{m}\)), fixed it in a specific shape, computed the shape and compared it with the segmented ground truth from the CT image. For the endovascular experiment, we inserted our fiber into a 3D printed vessel, which was created from a CT patient scan.

## Results and discussion

### Wavelength shift calculation

Table 1 shows the shape accuracies which were reconstructed with different reference wavelengths. The used wavelengths were determined experimentally by placing the fiber as straight as possible. For all three shapes, the obtained errors differ by several millimeters. This indicates that the reference wavelengths measurement has to be very accurate to obtain reconstructed shapes with a high accuracy. Thus, we recommend to determine the reference wavelengths by fixing it to a precise ruler, because here no kind of strain is applied.

### Strain computation

### Interpolation

*d*: a higher distance of the FBG arrays leads to more values that must be interpolated. For the distance of our fiber, the effects were low, see also [6].

### Curvature and angle computation

We calculated the curvature and direction angles using first a configuration with three cores (234), a configuration with three cores (347) and a configuration with four cores (2347). The numbers correspond to the cores as shown in Fig. 1, where the cores of configuration (2347) are highlighted in white. We observed that configuration (347) leads not to sufficient results due to the linear dependency of two cores. For proper 3D reconstruction, at least 3 linear independent cores are needed as observed with configuration (234) and (2347). The performance of FBGs in different cores in the multicore fiber varies quite significantly due to the manufacturing process. One side of the fiber will perform better, and this is why an asymmetrical configuration is preferable. Nonetheless due to low signals for shapes with high local bending radii, it is recommended to use more cores to ensure functionality for all possible shapes. Another possibility is to calculate multiple local strains for differing configurations and thus enable local averaging leading to higher accuracies.

### Curvature and angle correction

Results of the 3D experiment: measured errors \(e_{\text {avg}}\) and \(e_{\text {max}}\) in mm for different 3D shape measurements using our shape sensing model with and without the curvature and angle correction

Shapes\({\setminus }\)errors (mm) | With corrections | Without corrections | ||
---|---|---|---|---|

\(e_{\text {avg}}\) | \(e_{\text {max}}\) | \(e_{\text {avg}}\) | \(e_{\text {max}}\) | |

Circle 1 | 0.35 | 0.75 | 3.53 | 12.57 |

Circle 2 | 0.50 | 1.15 | 4.91 | 18.13 |

Circle 3 | 0.50 | 1.02 | 5.70 | 23.66 |

S-curve 1 | 0.70 | 1.29 | 5.77 | 17.83 |

S-curve 2 | 0.57 | 1.98 | 1.78 | 4.27 |

S-curve 3 | 1.15 | 7.53 | 2.33 | 6.57 |

Helix | 1.00 | 4.72 | 6.04 | 25.72 |

Inside the vessel | 1.13 | 2.11 | 2.59 | 7.00 |

### Shape reconstruction

### 3D shape reconstruction accuracy

For the 3D shape experiments, we integrated the results of all previous experiments in our model. We made several measurements bending our fiber to different 3D shapes. The segmented shapes from the CT scan of the measurements, as shown in Fig. 6, were used as ground truth. The accuracies, shown in Table 2, depend on the complexity of the forms: for the circular shapes, we obtain an average error of \({e_{\text {avg}} \approx 0.5~\mathrm{mm}}\) and a maximal error of \({e_{\text {max}} \approx 1~\mathrm{mm}}\), whereas for s-curved and helical shapes, we get higher errors, especially for *s-curve 3* and *helix*.

Comparing our results with Khan [7], we obtained higher errors. But Khan evaluated a catheter of only 114 mm length in different configurations with weakly bending, constant or linear curvature and nearly no torsion. Also they did not tested their catheter in any form with singularities like s-curves. Hence, the results of our 3D experiments with high deflections using the 380 mm multicore fiber are nevertheless accurate and promising.

## Conclusion

We presented an optimized shape sensing model with multicore fibers for flexible instruments. We conducted a detailed error analysis for every shape reconstruction step. The main error sources of shape sensing with multicore fibers are corrupted reference wavelengths for the wavelength shift computation, direction angles changed by the twist present in the fiber and curvature values, which are distorted by using an inaccurate photo-elastic coefficient or incorrect radii. This indicates that two calibration measurements need to be done for every fiber. The first one is used to determine the Bragg wavelength \(\lambda _\mathrm{b}\) with \(\kappa = 0\), the second one to get the twist angle \(\varphi _{\text {twist}}\) where \(\kappa \ne 0\) and the curvature scale factor. Further factors influencing the shape are the equations defined by the used core configuration, the curvature and angle interpolation and the reconstruction algorithm.

Furthermore, we evaluated the accuracy of our model with 3D measurements in a CT scanner. We received the accuracies \(e_{\text {avg}} \approx 0.35\) to \(1.15\,\mathrm{mm}\) and \(e_{\text {max}} \approx 0.75\) to 7.53 mm. Finally, we tested our fiber system in an endovascular scenario and obtained high accuracies (\(e_{\text {avg}}= 1.13 \,\mathrm{mm},\; e_{\text {max}}= 2.11 \,\mathrm{mm} \)). These experiments show promising results for using multicore fibers for shape sensing of catheters.

In future work, we aim to enable a full endovascular catheter navigation. For this purpose, we plan to combine the reconstructed shape obtained by the multicore fiber with the position and orientation of an electromagnetic tracking system. Also, the experiments were conducted under ideal conditions and we want to analyze our model in more realistic scenarios and to compensate other effects that may occur, such as dynamic change in the twist angle during an endovascular procedure. For this problem, using helically wrapped fibers as fiber system might be a better choice.

## Notes

### Acknowledgements

We thank Armin Herzog, Institute for Neuroradiology, University Hospital Schleswig-Holstein, Lübeck for the CT scanner support and the Department of Surgery, University Hospital Schleswig-Holstein, Lübeck for providing the 3D vessel printed by Fraunhofer EMB. This work was funded by the German Federal Ministry of Education and Research (BMBF, Project Nav EVAR, Funding Code: 13GW0228C).

### Compliance with ethical standards

### Funding

This work was funded by the German Federal Ministry of Education and Research (BMBF, project Nav EVAR, funding code: 13GW0228C).

### Conflict of interest

The authors declare that they have no conflict of interest.

### Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. This article does not contain any studies with animals performed by any of the authors.

### Informed consent

Informed consent was obtained from all individual participants included in the study.

## Supplementary material

## References

- 1.Banchoff TF, Lovett ST (2010) Differential geometry of curves and surfaces. CRC Press, Boca RatonGoogle Scholar
- 2.Black RJ, Zare D, Oblea L, Park YL, Moslehi B, Neslen C (2008) On the gage factor for optical fiber grating strain gages. In: Proceedings of society for the advancement of materials and process engineering (SAMPE ’08), the 53rd international SAMPE symposium and exhibitionGoogle Scholar
- 3.Cui J, Zhao S, Yang C, Tan J (2018) Parallel transport frame for fiber shape sensing. IEEE Photonics J 10(1):1–12CrossRefGoogle Scholar
- 4.Henken KR, Dankelman J, van den Dobbelsteen JJ, Cheng LK, van der Heiden MS (2014) Error analysis of FBG-based shape sensors for medical needle tracking. IEEE ASME Trans Mechatron 19(5):1523–1531CrossRefGoogle Scholar
- 5.Hill KO, Meltz G (1997) Fiber bragg grating technology fundamentals and overview. J Lightw Technol 15(8):1263–1276CrossRefGoogle Scholar
- 6.Jäckle S, Strehlow J, Heldmann S (2019) Shape sensing with fiber Bragg grating sensors. In: Bildverarbeitung für die Medizin 2019. Springer, pp 258–263Google Scholar
- 7.Khan F, Denasi A, Barrera D, Madrigal J, Sales S, Misra S (2019) Multi-core optical fibers with Bragg gratings as shape sensor for flexible medical instruments. IEEE Sens J 19(14):5878–5884CrossRefGoogle Scholar
- 8.Lee B (2003) Review of the present status of optical fiber sensors. Opt Fiber Technol 9(2):57–79CrossRefGoogle Scholar
- 9.Mendis S, Puska P, Norrving B (2011) Global atlas on cardiovascular disease prevention and control. World Health Organization, GenevaGoogle Scholar
- 10.Moore JP, Rogge MD (2012) Shape sensing using multi-core fiber optic cable and parametric curve solutions. Opt Express 20(3):2967–2973CrossRefGoogle Scholar
- 11.Ourak M, Smits J, Esteveny L, Borghesan G, Gijbels A, Schoevaerdts L, Douven Y, Scholtes J, Lankenau E, Eixmann T, Schulz-Hildebrandt H, Hüttmann G, Kozlovszky M, Kronreif G, Willekens K, Stalmans P, Faridpooya K, Cereda M, Giani A, Staurenghi G, Reynaerts D, Vander Poorten EB (2019) Combined oct distance and fbg force sensing cannulation needle for retinal vein cannulation: in vivo animal validation. Int J Comput Assist Radiol Surg 14(2):301–309CrossRefGoogle Scholar
- 12.Park YL, Elayaperumal S, Daniel B, Ryu SC, Shin M, Savall J, Black RJ, Moslehi B, Cutkosky MR (2010) Real-time estimation of 3-d needle shape and deflection for mri-guided interventions. IEEE ASME Trans Mechatron 15(6):906–915Google Scholar
- 13.Roesthuis RJ, Kemp M, van den Dobbelsteen JJ, Misra S (2014) Three-dimensional needle shape reconstruction using an array of fiber bragg grating sensors. IEEE ASME Trans Mechatron 19(4):1115–1126CrossRefGoogle Scholar
- 14.Rusinkiewicz S, Levoy M (2001) Efficient variants of the ICP algorithm. In: 3DIM, IEEE 145Google Scholar
- 15.Shi C, Giannarou S, Lee SL, Yang GZ (2014) Simultaneous catheter and environment modeling for trans-catheter aortic valve implantation. In: 2014 IEEE/RSJ international conference on intelligent robots and systems (IROS 2014), IEEE, pp 2024–2029Google Scholar
- 16.Werneck MM, Allil RC, Ribeiro BA, de Nazaré FV (2013) A guide to fiber bragg grating sensors. In: Current trends in short-and long- period fiber gratings. InTech, Rijeka, Croatia, pp 1–24 Google Scholar
- 17.Xu R, Yurkewich A, Patel RV (2016) Curvature, torsion, and force sensing in continuum robots using helically wrapped FBG sensors. IEEE Robot Autom Lett 1(2):1052–1059CrossRefGoogle Scholar

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.