Keywords

1 Introduction

In the United States, over five million central venous catheterizations (CVCs) are performed annually [6], with the internal jugular vein (IJV) being the most utilized insertion site [2]. The current gold standard is the blind technique, relying on the clinician to palpate surrounding anatomical structures to identify the insertion site (Fig. 1a). Ultrasound-(US)-guided CVC is becoming the preferred technique as it has the potential to reduce complications including accidental punctures to structures such as the carotid artery (CA) [7, 8]. The US-guided approach relies on real-time US video, depicting cross-sections of the anatomy on a 2D monitor, to guide the needle insertion (Fig. 1b). Despite US-guidance improving complication rates, clinical studies have found rates of CA puncture to be 7.8 % for US-guided trainees [5].

Fig. 1.
figure 1

Comparison of guidance techniques for CVC (a) Anatomical Guidance (b) US-only guidance and (c) AV guidance. Image (a, b) courtesy of Google

Full size image

The common use of US-guided interventions has resulted in development of many US-guided computer-assisted surgical navigation systems. To address the aforementioned limitations of US-guided CVC, an augmented virtuality (AV) monitor-based surgical navigation system was developed for needle guidance [1]. Their system employed magnetic tracking to render tracked virtual models of the US probe, needle and needle trajectory, onto a front-facing US image (Fig. 1c). This system did not demonstrate significant improvement in the complications associated with the needle insertion compared to the US-only technique for expert users [1]. Two potential factors that may have influenced the success of this system were the fixed face-on view provided to the user and the discrepancy between the visual and motor fields, as the user had to rely on a monitor exterior to the visual field of the phantom [1]. Despite the inconclusive results, their promising work has motivated our development of surgical navigation environments to reduce complications associated with CVC.

We investigated the efficacy of a first-person immersive mixed reality (MR) system for CVC needle navigation. Toward the long-term goal of clinical deployment, we first aim to understand how the method to visualize the surgical information affects the rate of complication during US-guided CVC. Our surgical navigation system combines a spatially tracked head-mounted display (HMD) system with a surgical magnetic tracking system, allowing magnetically tracked surgical instruments to be visualized in 3D inside the HMD with submillimetre accuracy. For this work, we compare US-only guidance to a MR guidance system displayed on a 2D monitor or within a HMD. We hypothesize that the HMD will improve the success of needle insertions compared to the US-only and 2D monitor systems. This work aims to highlight the importance of coherent visual and motor fields for surgical applications. Aside from the technical advancement, our contribution is a comprehensive user study involving 33 clinical practitioners.

2 Materials and Methods

A patient-specific neck vasculature phantom was constructed, comprising of a hollow (wall-less) vascular structure embedded in an US compatible solid medium [9] (Fig. 2a). The positive models of both the CA and IJV were manufactured using 3D printer based on the manual segmentation of a patient computed tomography (CT). These 3D printed vessel models were embedded into, and later removed from, the US-compatible medium, leaving the exact negative imprint of vessel geometries. The phantom was housed in a plastic container with 8 hemi-spherical fiducial markers and scanned in CT (O-Arm, Medtronic, USA). The segmentation of the vessels and fiducial markers served as the basis for visualization and registration with the tracking system. This phantom serves as a surrogate for patient anatomy, producing anatomically realistic US images (Fig. 2b) compared to those obtained from healthy volunteers (Fig. 2c and d).

Fig. 2.
figure 2

(a) Phantom development, (b) Phantom under US, (c) and (d) health human neck vasculature under US. Image (c, d) courtesy of health volunteer.

Full size image

The surgical guidance system comprises of a monitor or mixed-reality (MR) HMD (HTC VIVE Pro, HTC, Taiwan ROC), a magnetic tracking system (Aurora, NDI, Canada), a clinical US scanner (SonixTouch, BK Medical, USA), and a surgical hypodemic needle (7 cm metallic needle with 10 ml syringe, Fig. 3b). A linear transducer (L14-5,BK Medical, USA) was used to acquire real-time images of the phantom (depth of 6 cm with gain of 42 %). The US transducer, neck phantom, and surgical needle were magnetically tracked, spatially calibrated, and registered into a common coordinate system. As the HTC VIVE Pro has its own tracking system, it is co-registered with the magnetic tracking by means of a co-tracked apparatus that registers the magnetically tracked tools into the HMD (Fig. 3a) [4]. The co-calibration method was previously validated using a Computerized Numerical Control machine with reported accuracy of less than 1 mm and 1\({}^{\circ }\) (trueness plus precision) [4]. The magnetically tracked US transducer was calibrated using a Procrustean method [3], and the surgical needle was calibrated using a template-based approach (Fig. 3b). A visual assessment of the system accuracy, comprising of trackers co-calibration, tool calibration and patient registration is shown in Fig. 3c.

Fig. 3.
figure 3

(a) The co-calibration apparatus tracked by the VIVE controller and magnetic pose sensor (b) the calibration apparatus for the syringe, and (c) Visual representation of the total system accuracy

Full size image
Fig. 4.
figure 4

Visual representation of each mode of visualization where (a) is the US-only system, (b) is the monitor system, and (c) is the HMD system. Images (b) and (c) comprise of models of the US probe, needle, needle trajectory and the calibrated US image.

Full size image

Three modes of visualization were implemented and evaluated: (1) the traditional US-only, displayed on the US scanner, (2) an AV system displayed on a 2D monitor, (3) a MR system displaying the streaming US and tracked tools in their registered and tracked pose using the HMD, as depicted in Fig. 4. In both AV and MR visualizations, streaming US video, virtual representation of the tracked probe and needle, as well as a needle trajectory represented as a 10 cm blue extension from the needle tip were displayed for surgical guidance.

Thirty-three expert clinicians were recruited with consent according to the local REB regulation (Western University REB 107254). To accommodate hospital scheduling, the study was designed to take be 15-minutes in duration. Prior to the experiment, each participant was briefed on and introduced to the needle insertion required for CVC using the neck phantom. The vasculature on the left side of the neck phantom was used to train the users on all of the systems. The participants were given time to perform needle insertions using the US-only, MR on a monitor, and MR in the HMD system, until the user felt they were comfortable with all of the modes of visualization. The study was conducted using the vessels on the right hand side of the phantom. The order for each set of insertions was randomized for each participant. The participant was required to perform one insertion into the vessel on the right hand side of the phantom for each of the modes of visualization. Sufficient time was provided in between switching modes of visualization to allow the participant to rest and adapt to the new mode. The streaming US video, time, and tracked trajectories were recorded. After performing the experiment, the users filled out a questionnaire.

3 Results

Fig. 5.
figure 5

(a) Successful insertion (b) Needle tip position exterior to the IJV (c) CA puncture (color figure online)

Full size image

The recorded data were processed to produce the following metrics: time, distance from the final needle tip position to the vessel wall, and path length. The recorded tracking information for all tracked apparatus were used to generate the needle path, visualized as heatmap-coded spheres in Fig. 5. Red spheres indicate the beginning of the needle insertion, transitioning into cold colour as time progresses. The needle path relative to vascular structure was analyzed to extract: the number of successful insertions define by insertions with the final needle tip within the IJV with no CA punctures (Fig. 5a); the number of insertions where the needle tip ended exterior to the IJV (Fig. 5b); and the number of CA punctures (Fig. 5c) . The questionnaire responses were in the form of a continuous scale where the centre and two ends were anchored with written descriptions. If the user agreed with one of the given responses they could mark that part of the scale or alternative anywhere along the scale. The questionnaire responses were converted into a numeric 10.0 scale and summarized in Table 1.

Table 1. User questionnaire results
Full size table
Fig. 6.
figure 6

(a) Graphical depiction of the number of successful insertions performed for each of the guidance systems, and (b) Graphical depiction of the average distance from the final needle tip position to the vessel wall for each of the guidance systems.

Full size image

The continuous results, such as time, distance to vessel wall and path length, were individually compared across the three conditions US-only, monitor, and HMD using a repeated measures ANOVA. The ANOVA for the distance from the vessel wall returned a p-value less than 0.05 and therefore a least squared distance mutli-comparison post-test was conducted to compare between each pair of conditions. This post-test returned the p-value 0.044 for the US-only and HMD conditions indicating significant differences between these two conditions for the distance from the vessel wall. The mean and standard deviation of the distance from the vessel wall were calculated for each conditon, as the ANOVA for this metric returned a significant p-value. For the discrete results, including the number of successful insertions and CA punctures, the McNemar test was performed. Using this test to compare the number of successful insertions between the US-only and HMD returned a p-value of 0.0106 indicating a significant result. Therefore, the total number of successful insertions for all conditions was calculated. The number of successful needle insertions and distance to the vessel wall have been summarized in Fig. 6, with significant combinations (p < 0.05) denoted. All metrics that were not significant were not reported.

4 Discussion

The HMD system significantly improved the number of successful insertions and increased the distance from the final needle tip position to the vessel wall such that the needle tip was more centred within the lumen of the vessel compared to the US-only system. Thirty-one of the thirty-three clinicians performed successful insertions using the HMD system, whereas only twenty-one performed successful insertions using the US-only system. The HMD system had an average distance to the vessel wall of 3.8 ± 3.1 mm, whereas the US-only system had an average distance of 2.2 ± 4.4 mm. If the needle tip was within the vessel lumen the distance was denoted as positive, whereas if the needle tip was exterior to the vessel the distance was denoted as a negative. Therefore, a large positive number is desired as it is representative of the needle tip position within the vessel but far from the vessel wall. Clinicians using the HMD system produced a significantly larger distance on average with a smaller standard deviation compared to the US-only system. Thus, on average, clinicians more consistently targeted the centre of the vessel with the HMD system compared to the US-only system. However, other metrics that were calculated, such as path length and time, were not significantly different between any of these conditions. While these metrics did not show significant improvement, the HMD system allowed for improved guidance compared to the US-only and monitor systems without having a significant impact on the insertion time or path length.

There was no significant improvement in the clinicians’ needle guidance when using the monitor based system compared to the US-only system. This result is consistent with the work done by Ameri et al.[1], where a similar monitor based system showed no significant improvements for expert users’ needle guidance compared to an US-only system. The lack of improvement is likely due to disparities between the clinicians’ motor and visual fields when using both the US-only and monitor based systems. Despite the fact that the monitor system had additional information intended to improve the needle guidance, clinicians’ performance was comparable between the US-only systems. However, the HMD system did have significantly improved guidance compared to the US-only system. These results emphasize the importance of coherent visual and motor fields of needle guidance, which can be accomplished using a HMD. While monitor based systems would be more simple to integrate into a clinical workflow, the results presented here suggest that using HMD to bring needle guidance information directly into the line of sight of the clinician is important for successful needle guidance, promoting the continued pursuit of HMD research.

The rates of CA puncture are low and were not significant and therefore are not reported in this paper. The lack of CA punctures is likely due to the simplicity of the phantom as the IJV and CA had a simple orientation with limited overlap, as the CA was positioned laterally to the IJV. However, as depicted in Fig. 2(c and d), the appearance and configuration of the human neck vasculature is variable. The number of insertions that resulted in the final needle position external to the IJV could be representative of the potential puncture risk to important adjacent anatomical structures due to this variability of the anatomy. Therefore, the reduction of the number of final needle placements outside of the IJV using the HMD could be a surrogate for a reduction in overall complications.

While this study supports the use of HMDs for needle guidance, it is important to consider the feasibility of clincial integration. Questionnaire responses showed that the current HMD system used for this study may be more useful for training rather than clinical use, as on average clinicians ranked the clinical viability of the system a 4.35 out of 10 compared to a 7.04 for usefulness for training. The centre of the scale (5/10) for the question on clinical viability represented the response “viable with proper assistance”. Thus, on average, clinicians reported that with assistance this technology could be used in the operating room. However, on average, more clinicians see the potential of this system as a training tool. Furthermore, on average clinicians (4.75/10) indicated they would use this technology on a case by case basis, as more complicated cases would benefit from this high level of guidance. Future work will include continued development of the HMD system to include more real-world information to improve clinical feasbility. This study was performed entirely in VR to isolate the effects of the visual and motor fields. However, many of the clinicians did not think the current system was fully clinically viable due to the lack of real world information. Integrating feed from stereo camera would allow the clinician to visualize the guidance information and the real world simultaneously. Alternatively, the monitor based system is more clinically feasible due to the similarity to current US-only guidance platforms.

5 Conclusion

We developed an advanced needle guidance system that renders tracked tools such as the US probe, needle, and needle trajectory as well as the calibrated US image on a 2D monitor or within a HMD. The aim of this research was to compare needle insertion performance using US-only guidance, the advanced guidance system on a monitor, and the advanced guidance system in the HMD. Thirty-three expert users were trained on all three systems and then used each system to perform a needle insertion on the phantom. The HMD system significantly improved the number of successful needle insertions, as 31 out of 33 clinicians had a final needle tip position within the vessels lumen compared to 21 out of 33 clinicians using US-only guidance. The HMD system also significantly improved the distance from the final needle tip position to the vessel wall. Clinicians using the HMD system had an average distance of 3.8 ± 3.1 mm compared to 2.2 ± 4.4 mm using the US-only system, meaning they were consistently closer to the centre of the lumen of the vessel compared to the US-only approach. The monitor system did not show any siginficant improvements compared to the US-only system. Therefore, using a HMD to align the visual and motor fields is an important factor in promoting successful needle guidance, encouraging the continued pursuit of HMD surgical navigation research.