Oriented 3D Ultrasound for Central Venous Cannulation Using an Augmented 2D Ultrasound System.

Abstract

Two-dimensional ultrasound (2DUS) guidance for central venous cannulation is a best practice for patient safety identified by the Agency for Healthcare Research and Quality and the American College of Emergency Physicians.1 2DUS has been shown to reduce mechanical complications of insertion (e.g., arterial puncture, pneumothorax) by up to 71% and improves procedure performance (e.g., reduced total procedure time, reduced number of skin punctures, decreased number of needle redirection events), but complications remain in the setting of ultrasound guidance, including pneumothorax and arterial puncture.2 Puncture of the posterior vessel wall (a surrogate for incorrect needle tip placement and a potential contributor to hematoma formation and inadvertent arterial puncture) occurred in 34% of ultrasound-guided central venous catheter insertions in a phantom study, including 31% with short-axis (SA) approach and 37% with long-axis (LA) approach (no statistical difference).3 The internal jugular vein (IJV) overlies the common carotid artery (CCA) in approximately 50% of patients, raising concern that posterior wall puncture (PWP) of the IJV could lead to puncture of the CCA.4 The limitations of 2DUS in preventing mechanical complications arise from the fact that users must select from one of two common approaches, each with advantages and disadvantages. SA (also called “out-of-plane”) approach allows the user to center on the target vessel of interest, but the needle tip is commonly not visualized, and the depth of insertion cannot be judged accurately. If the needle tip beyond the image plane deviates medially, laterally, or deep to the vessel, this cannot be identified without active tracking of the needle by the user. LA (also called “in-plane”) approach allows the user to see the entire length of the inserted needle in a single image. The depth of insertion can be seen at all times, reducing the risk of the needle being inserted too deeply. However, skill is required to align the transducer, ultrasound plane, needle, and target, and the adjacent artery is not seen. Errors have been described with this method, including PWP of the target vein and inadvertent puncture and even cannulation of the adjacent artery.5 A phantom study comparing LA and SA approaches showed shorter time to cannulation, fewer needle redirections, and lower rates of PWP with LA, compared with SA.6 Even with LA approach, PWP rates of 21% were observed for IJV cannulation. Combined SA and LA approach (skin puncture with SA visualization, followed by vessel puncture under LA visualization) in a phantom model allowed higher procedure success rate for novices and reduced rate of PWP in a nonrandomized study in patients.7, 8 However, a randomized study of patients undergoing cardiac surgery showed higher first-pass success rates and fewer needle redirections with the SA approach, compared with LA, and carotid artery puncture occurred only in the LA approach group.5 Complications might be mitigated by the simultaneous visualization of LA and SA planes or by the depiction of a three-dimensional field of view of interest as a volume-rendered model. Existing three-dimensional ultrasound (3DUS) systems are typically designed for cardiac or fetal visualization and thus do not include high-frequency transducers, which are commonly used to guide catheter insertion because of their high resolution for superficial structures. Most traditional 3D-capable systems are also expensive and large, making them less well suited or less feasible for the bedside insertion of catheters in emergency and acute care environments. A low-cost system for augmentation of any 2DUS system and transducer to generate automatically oriented 3D volumes can be constructed using an inertial measurement unit (IMU).9, 10 We hypothesized that an IMU-augmented 2D high-frequency linear ultrasound transducer would enable rapid 3DUS visualization of the IJV and carotid artery as well as needle, wire, dilator, and catheter. We tested this in vivo and in a phantom model. We paired a 2D linear high-frequency transducer (FujiFilm SonoSite HFL38) and ultrasound system (FujiFilm SonoSite M-Turbo) with an augmentation system incorporating an IMU (STEVAL-MKI121V1INEMO-M1 Discovery, STMicroelectronics).9, 10 Source 2DUS images were captured from the video-out port on the ultrasound system using a capture device (Epiphan DVI2USB 3.0). A computer (Dell Precision M4800) was used to perform image reconstruction using software designed by the research team (C++, MatLab). Reconstruction was triggered automatically, immediately following image acquisition. Total acquisition time (including transducer fanning and file writing), total reconstruction time (including MatLab and data loading, image filtering, reconstruction, and file writing), and angular field of view were automatically logged by the software and are reported as measures of clinical feasibility of the technique. During acquisitions, the operator fanned the IMU-equipped ultrasound probe once around a central rotational axis on the phantom or body surface to traverse the region of interest, spanning an arc of approximately 70°. Source 2D images were transformed to 3D volumes using a pixel-based reconstruction algorithm, with a slice thickness of 5 pixels. Volumes were visualized using 3DSlicer (https://www.slicer.org/), an open-source medical image application. Using a central venous cannulation model (BluePhantom, CAE), images were acquired before insertion and at each stage of insertion of a central venous catheter (Arrow Multi-Lumen central venous catheterization kit, Teleflex; needle inserted, guidewire inserted, dilator inserted, catheter inserted). Images of the IJV and carotid artery were also acquired in a single human subject. Institutional review board approval was obtained for the study. Informed consent was obtained from the human subject. Representative images of the IJV and carotid artery in the human subject are shown in Figure 1. The region of interest is shown as a volume-rendered image (left) and planar stacked images matching the cardinal anatomic planes (right). The human figure icon is an automated orientation feature that rotates with the image volume, as do letters indicating anterior (A), posterior (P), right (R), left (L), inferior (I), and superior (S). Additional images of each of the stages of catheter insertion in the phantom model are available for online review (Data Supplement S1, Figures S1–S5, available as supporting information in the online version of this paper, which is available at http://onlinelibrary.wiley.com/doi/10.1111/acem.13831/full). The mean angular field of view was 73.6°. Image acquisition time averaged 18.31 seconds. Mean reconstruction time was 12.0 seconds. Using an IMU-augmented 2DUS system, we demonstrated 3D visualization of the IJV and carotid artery in a human subject and in a simulated phantom model of central venous catheter insertion. Field of view was sufficiently broad to include relevant anatomy and medical devices. The resulting 3D volumes can be flexibly rendered, including windowing methods similar to CT, facilitating visualization of anatomy and medical equipment. A variety of other procedures might benefit from this imaging technique, including any procedure requiring precise and accurate insertion of a needle or catheter into a target while avoiding adjacent vulnerable nontargets. Examples include peripherally inserted central catheter placement, peripheral intravenous catheter insertion, arterial cannula insertion, abscess drainage, percutaneous nephrostomy, cholecystostomy, catheter-directed ablation of tumors, percutaneous needle biopsy of masses, regional anesthesia catheter placement, stenting of ducts and vascular structures, and amniocentesis. Benefits include low cost and wide compatibility with existing ultrasound systems and transducers, allowing use of existing equipment. This was a preliminary study of feasibility and cannot determine clinical utility. Although acquisitions and reconstructions were rapid, we did not explore image quality resulting from even shorter acquisitions and reconstructions, which would be desirable for real-time guidance of certain procedures. Total acquisition time plus reconstruction time averaged approximately 30 seconds; improvements in speed would be desirable for critical care scenarios. The minimum (fastest) acquisition time providing sufficient image quality for recognition of anatomy and medical devices should be determined in future studies. Because image reconstruction time is linearly related to acquisition time (via the number of acquired 2D pixels requiring reconstruction into corresponding 3D voxels), shortening acquisition time would also improve reconstruction time, even in the absence of other algorithm improvements. Image reconstruction could be performed in parallel with acquisition, rather than sequentially as in the current approach. Under our algorithm, reconstruction time is uniformly shorter than acquisition time; therefore, if performed in parallel, the reconstructed image would be available almost instantly upon completion of acquisition. Faster processors and cropping or down-sampling source images to reduce pixel counts requiring reconstruction are additional means to enhance speed. Future studies using appropriate blinded comparisons are needed to determine whether 3D visualization reduces clinically relevant procedure complications, preserves important feasibility factors such as procedure speed, and achieves acceptable image quality in comparison to standard 2D techniques. We successfully imaged a central venous catheter insertion using 2DUS augmented to produce 3D images. Field of view was broad and imaging times totaled approximately 30 seconds. Additional work is needed prior to broad clinical adoption. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.