Active Ultrasound Pattern Injection System (AUSPIS) for Interventional Tool Guidance

Accurate tool tracking is a crucial task that directly affects the safety and effectiveness of many interventional medical procedures. Compared to CT and MRI, ultrasound-based tool tracking has many advantages, including low cost, safety, mobility and ease of use. However, surgical tools are poorly visualized in conventional ultrasound images, thus preventing effective tool tracking and guidance. Existing tracking methods have not yet provided a solution that effectively solves the tool visualization and mid-plane localization accuracy problem and fully meets the clinical requirements. In this paper, we present an active ultrasound tracking and guiding system for interventional tools. The main principle of this system is to establish a bi-directional ultrasound communication between the interventional tool and US imaging machine within the tissue. This method enables the interventional tool to generate an active ultrasound field over the original imaging ultrasound signals. By controlling the timing and amplitude of the active ultrasound field, a virtual pattern can be directly injected into the US machine B mode display. In this work, we introduce the time and frequency modulation, mid-plane detection, and arbitrary pattern injection methods. The implementation of these methods further improves the target visualization and guiding accuracy, and expands the system application beyond simple tool tracking. We performed ex vitro and in vivo experiments, showing significant improvements of tool visualization and accurate localization using different US imaging platforms. An ultrasound image mid-plane detection accuracy of ±0.3 mm and a detectable tissue depth over 8.5 cm was achieved in the experiment. The system performance is tested under different configurations and system parameters. We also report the first experiment of arbitrary pattern injection to the B mode image and its application in accurate tool tracking.


Introduction
Tool tracking/guidance techniques have broad utility in medical intervention and image-guided therapy. In medical interventions, including biopsy, ablative therapy, and cardiac catheterization, image guidance is the pivotal task and technology that determines operation success and guarantees clinical outcome. In literature, research shows that in many applications, intervention outcome can be greatly improved if an effective tool tracking method is introduced [1]. Various medical imaging modalities are used to recover the pose of the interventional tools [2]. Compared to CT and MRI, ultrasound (US) imaging has several advantages including low cost, safety, mobility and ease-ofuse. Two major limitations, however, prevent conventional ultrasound imaging systems from providing effective tracking and guidance. The first one is the poor visualization of interventional tools. Ultrasonography reconstructs images from received echo ultrasound signals. In interventional procedures, many catheters or needles cannot be reliably visualized in ultrasound images due to the low echo signal amplitude, large reflection angle, strong tissue attenuation or the complicated image background [3]. The second limitation is the low localization accuracy in the elevational dimension. The acoustic beam width usually ranges from several millimeters to one centimeter, which means an object shown on a 2D ultrasound image has an inherent uncertainly of a few millimeters to one centimeter along the elevation axis.
In recent years, several approaches have been proposed to enable tool visualization and pose recovery, including beam steering, optical tracking, electromagnetic (EM) tracking, and passive/active ultrasound markers [4][5][6][7][8][9][10][11]. The beam steering method has been developed and validated by several research groups and currently is integrated into a commercial ultrasound scanner from SonoSite Inc. The basic principle is to steer the imaging beam directions to get the optimized reflection from the catheter. It has been proven effective when the catheter is in-plane of the US image. For cases where catheters simply intersect the imaging plane with a large angle, this method is not effective [4]. In the optical tracking approach, an optical marker is attached at the end of the tool and uses a stereo camera to capture the 3D position of the marker [5]. This approach suffers from several limitations, including high cost, calibration complexity, intrusive setup, needle bending error, and the need for clear line-of-sight from the camera to the tool. Electro Magnetic (EM) tracking approach was recently introduced and integrated into several commercial ultrasound scanners (GE LOGIQ E9, Ultrasonix GPS, etc.). In this approach, an EM field is generated by an EM emitter; EM sensors are implanted in both the catheter and the imaging probe. The relative pose of the catheter is estimated and injected as a graphic overlay to the B-mode image [6]. The EM tracking quickly became a commercialized technology due to its merits of excellence in off-plane detection capability. However, several disadvantages limit its applications to all interventions: the overall navigation accuracy can easily be worse than 2,3 mm; it requires specially designed imaging equipment; and any ferromagnetic object in the operation region may affect the system accuracy. Another approach with a similar concept is to use ultrasound sensors instead of EM sensors. Several research groups have demonstrated the system in which an ultrasound receiver is attached to a catheter. The catheter localization is retrieved by measuring the timing of the ultrasound pulses from the imaging probe [8][9][10]. This method requires specialized US system and image processing software; thus its application is limited.
Other than detecting the tool spatial location, some researchers focus on the visualization enhancement of the catheter in B-mode images. One approach is to improve image quality by using passive ultrasound markers. In this method, ultrasound markers, sometimes in the form of scattering coatings, are integrated into the tool to improve the echo amplitude [7]. The major issue of this approach is that the visualization enhancement by scattering coating is limited; with markers, they are usually too bulky for small catheters. On the other hand, active ultrasound marker approach has also been proposed. F. Simonetti introduced the concept that using biopsy needles as ultrasound waveguide [11]. An ultrasound wave is injected from the end of the needle. It travels along the needle and leaks into the tissue, and is picked up by an external imaging probe working in passive mode. The limitation of this method is that the phase, waveform, timing and leaking position of the guided beam cannot be well controlled; It requires the ultrasound system to stop normal operation and work in passive mode, which is not a direct visualization enhancement; its performance highly relies on the mechanical property of the needle and cannot be generally applied to any surgery tools. B. Breyer and I. Cikes reported an active ultrasound marker system that fires an ultrasound signal from the catheter, which shows a visible pattern in the B-mode image [12]. The timing is acquired by receiving the imaging pulses from the ultrasound probe. This approach directly injects the marker in the B-mode image and does not require special designed imaging system. B. Breyer's and I. Cikes' work for the first time presented the active marker concept on the medical ultrasound imaging system. However, this method was not eventually used in clinics. Experiment results show that in the in-vivo B mode images, it is difficult to identify the active echo spot from the tissue texture. Even in the water tank environment, a simple receive and echo configuration does not provide enough feedback to the operator; the echo spot may show up within a large range of tool position, the elevation localization accuracy is poor.
In this paper we introduce a new technique called ultrasound active pattern injection, a further development of the active visualization enhancement approach. Unlike previous works, the tool visualization and localization is not simply achieved by introducing an ultrasound beam to the tissue or echoing the imaging pulses. The Active Ultrasound Pattern Injection System (AUSPIS) is composed of an interventional tool with one or multiple ultrasound elements (active echo element), ultrasound analog frontends, a signal processing system, and a pulser. The system receives the image beacon pulses, analyzes the acquired signal, and fires one or a series of active echo pulses from the same active echo element with a proper timing, frequency, duration and amplitude. Thus it enables us to inject any ''virtual'' pattern into the B-mode image. Unlike the image overlay technologies used in other tracking or image registrations, the pattern injected to the B mode image from the actively encoded ultrasound field in the tissue. The encoding is based on the B-mode ultrasound image forming mechanism, so it doesn't require any hardware or software modification to the ultrasound machine. It continuously measures the local acoustic signal amplitude, by which a submillimeter elevation localization accuracy can be achieved. By configuring the pattern formations, the technique can be used for tracking, tool guiding, and annotation. In this study, we 1) developed a hardware and firmware prototype of the AUSPIS; 2) performed the lab bench experiment to test the design idea and investigated the working conditions and limitations of the system; 3) tested the system performance in-vivo.

Methods
The basic principle of AUSPIS is shown in figure 1. When acquiring a B-mode image, one or a group of array elements is driven to fires an ultrasound pulse, which propagates along a transmission beam line. After the pulsing, the array is switched to receive mode to acquire the echo. Element groups are fired one by one until the entire image region is covered to reconstruct ultrasound Brightness mode (B-mode) from the multiple RF lines. In the active pattern injection system, a small ultrasound transducer (active echo element or AE element) is integrated with the catheter. When a transmission beam reaches the element, an electrical pulse signal is excited and sent to the electronics. To simply improve the tool visualization, the electronics drive the AE element immediately to send an ultrasound pulse back to the imaging probe. The receiving-transmitting delay is in nanoseconds and negligible for US imaging. The active ultrasound pulse is superimposed on the catheter echo wave, resulting in an enhanced echo pulse with a much higher amplitude, broader frequency range and wider emission angle. This pulse travels back to the imaging probe, and appears as a bright spot (AE spot) that indicates the AE element location in the B-mode image.
As described previously, simply echoing the imaging pulses is not very useful in many cases. However, based on the signal received by the AE element, more complex functions can be developed to improve the tracking performance.

Time modulation
In clinical or ex-vivo conditions, simple active echoing has very limited improvement in the visualization, because a static echo spot can barely be seen from a complex B-mode background. To overcome this problem, the signal processing unit periodically enables/disables the active echo feedback. As a result, the active echo spot blinks in the B mode image with the same period, making it distinct from the background B mode textures. The tissue phantom and in-vivo experiments proved a good improvement of catheter visualization and tracking using this method.

Frequency modulation
The signal processing unit can also control the echo pulse frequency and duration. Although the signal waveform is not shown in the B-mode image, in the raw ultrasound or beam-formed RF signals, filtering method can be used to differentiate the background B-mode signal and the active echo signal, which has a different frequency or waveform characters than the imaging pulses. This can be used for the automatic tool guidance.

Mid-plane detection
In surgical interventions, even if the tool shows good visualization in B-mode image, it is difficult to determine whether it is well aligned with the image mid-plane due to the ultrasound beam width, as was previously described. To achieve high localization accuracy, a tracking system should be able to detect the image mid-plane, as shown in figure 1. In ultrasound imaging, the beam transmitted from an imaging probe has higher power intensity at the center, and lower intensity away from the center. Since the AE element is capable of measuring the local ultrasound pressure in real time, the beam intensity distribution can be utilized to localize the mid-plane. In the AUSPIS, when the AE element is well aligned with the mid-plane, the received beacon signal amplitude reaches its maximum, and when it moves away, the strength of beacon is decreased. Practically, the US beam intensity is determined by many factors including probe type, tissue attenuation, image depth, focusing point, etc. It is difficult to preset a trigger threshold that makes the AUSPIS only respond to the maximum US signal. However, we found that if the threshold is set to a relatively low value, the AUSPIS will be triggered multiple times in each image frame, and the trigger count, which is defined as the number of triggers received in each frame, reaches its maximum when the AE element is well aligned with the midplane. This is because when the US probe scans the FOV, each beacon beam has spatial overlapping with one or more of its neighboring beams. So the AE element receives US signal not only from the beam that intersects it, but also from the nearby beams, just with lower amplitude. When the AE element moves closer to the mid-plane, the received signal amplitude increases; as a result, more nearby beams exceed the threshold, so the trigger count increases. Since the trigger count indicates the alignment between AE element and mid-plane, it can be used for mid-plane detection. AUSPIS is able to read the trigger count and feedback to the operator in multiple ways, like a number displayed on the screen, a beeping sound with a frequency proportional to the trigger count, or an injected virtual pattern on the B-mode image, which will be discussed later.

Arbitrary pattern injection
The pattern injected into the image is not limited to the AE spot. As described previously, a B-mode image is formed by a series of A-mode lines. As shown in figure 2, suppose a B mode image is composed of 16 A-mode lines; the normal distance from the AE element to the probe is y. To generate a virtual spot on the point A, an ultrasound pulse should be received by the probe when the A-mode line #4 is being acquired, with a delay of where c is the speed of sound in this medium. The ultrasound pulse is generated from the AE element at position O, the distance between O and the center of the imaging elements R is d. The time for sound to travel from O to R is t travel = d/c. So the timing that the element should send an ultrasound pulse is: In other words, if an ultrasound pulse is fired from the AE element t seconds after the probe starts acquiring the A mode line #4, it will be shown as a virtual spot at position A in the B-mode image.
Having these spots as ''pixels'', arbitrary patterns can be formed and injected to the image. The challenging issue is that the relative position between O and R is unknown in the real application scenario. Two methods can be used to overcome this problem. The first method can be used if the ultrasound system is able to output synchronization signals. Since the linear array size, the total number of A-mode lines, and the speed of sound are known parameters, d can be automatically determined from the delay between received signal and the synchronization signal. In this case the arbitrary pattern can be injected into any pre-defined position in the B-mode image, i.e., the injected pattern can be either at an absolute position in the image coordinate, or at a relevant position moving with the active element. The second method can be used when the synchronization signal is unavailable. In this case injecting a pattern to an absolute position in the image coordinate is not possible. However, since the AE element is able to receive multiple neighboring beacon pulses when the probe is acquiring the nearby A-lines, the A-line period can be acquired by the system. Using the received pulse as the synchronization signal, and calculating the desired time delay from the A-line period, virtual patterns can be injected into the B-mode image with the AE element coordinate. The implementations of both methods will be presented in the following paragraphs. Figure 3a shows the block diagram of the AUSPIS prototype. A piezoelectric element for both ultrasound receiving and transmitting is integrated with a catheter. The element is connected to a customized electronic circuit, which consists of a transmit/receive (T/R) switch, a variable gain amplifier (VGA) with analog filters, triggering circuit, analog to digital converter (ADC), an embedded microprocessor, and a pulser. Figure 3b and 3c show the circuit board and the catheter tip. The T/R switch circuit is built with a TX810 8 channel integrated switch to protect the receiving circuit from the transmission driving voltage, which can go up to a hundred volts. Analog filters have a low and high cutoff frequency of 0.1 MHz and 20 MHz. The trigger circuit compares the absolute value of the signal amplitude and the threshold, and sends  pulses out when it is higher than the threshold. Since the receiver and emitter are the same element, the trigger circuit has a latch and reset function to prevent oscillating triggering. The ADC converts the analog signal to digital. The control program runs on an embedded processor, which controls all the peripherals on the AUSPIS circuit. To drive the element, the pulser is able to generate pulses with a minimum duration of 12.5 ns and variable voltage from zero to 6150 V. The active echo element is a customized small tube made of PZT5-H material with an outer diameter of 2.08 mm, inner diameter of 1.47 mm, and length of 2 mm. The catheter has a sealing layer around the element, which makes the overall diameter of the catheter tip close to 3.0 mm. A Sonix RP and a Sonix Touch ultrasound systems (UltraSonix Co.) with L14-5W (128 element) linear probe and 4DL14-5/38 (128 element) 3D probe is used in the experiment. We've also tested the AUSPIS with SonoSite portable ultrasound system to verify the cross platform operation performance. Figure 4a indicates that the prototype device has an overall electronic loop delay of 0.6 ms. Due to this delay, the active echo spot is about 0.3 mm away along the axial direction from the true element position in the B-mode image. Figure 4b shows the signal received by the imaging probe. The red waveform is the reflection from the catheter; the active echo system is not enabled. The blue line shows the received waveform when the active echo is enabled.  From this plot we can clearly see that the active echo is fired after the reflection signal, with a delay of around 1 ms. The additional 0.4 ms delay may be due to the response of the prototype PZT element, which is relatively large in size. The active echo adds a ringing tail after the original signal. The frequency of the active echo is adjustable; in this plot, it has a slightly higher frequency than the original signal.

Mid-plane detection
As discussed previously, either the peak signal amplitude or the trigger count can be used in mid-plane detection. Practically, in the trigger count method the AUSPIS starts to response the beacon pulse even if the AE element is not well aligned with the mid-plane, it provides a large detection range, which is useful for a quick target search before the accurate tool tip localization. Once  the active echo signal is received, the peak signal amplitude method can be used for a more accurate alignment.
A typical trigger count versus AE element to mid-plane distance plot is shown in Figure 6. The data is acquired with a fixed AUSPIS receiver gain of 19 dB. The AE element is about 4 cm away from the probe. The US probe is running on its maximum transmission power with 32 element transmission aperture and 4 cm focus depth. In each frame the probe fires 256 RF lines. When the offset increases, the frame trigger count drops from the maximum 41 to 0 at an offset of 8 mm. The result indicates that, with these system parameter settings, the active echo detectable range is about 68 mm from the mid-plane. The range can be increased or decreased by using higher or lower amplifier gain.
Once the AE element is triggered, the next step is to accurately align it to the mid-plane. Figure 7a, b shows the catheter received signal amplitude verses the imaging probe position. In the plot, the mid-plane position is indicated by the maximum of the signal amplitude. If the trigger level is set to a value close to the maximum, the active echo system can only be triggered when the element overlaps the mid-plane. For example, if a trigger is set to 90% of the maximum amplitude, a localizing accuracy of 60.5 mm is expected from the Figure 7. In the water tank validation, the receiver gain is decreased until only one trigger per frame is achieved at the well aligned position. Then gradually move the catheter away from the mid-plane. At an offset of 60.31 mm, the active echo spot disappears. The result is repeatable and stable over test runs. Using this method, a midplane localization accuracy of 0.3 mm can be achieved.

Insertion angle test
In the previous tests, the catheter is perpendicular to the imaging ultrasound beams. However, this condition is not always satisfied in clinical applications. An experiment is performed to validate the system performance when the catheter is inserted at an angle. Since the prototype active element has a cylindrical shape, the most sensitive configuration is the cylinder axis perpendicular to the ultrasound beam, which is marked as 0u in the experiment. Keeping the element at the same location, the catheter rotates from 250u to 90u with 10u increment steps. Because the element is not an omnidirectional sensor, the received ultrasound pulse amplitude varies with the angle, so the receiver gain needs to be adjusted at different angles. The active echo performs properly at all angles we tested in the experiment. Because the catheter interferes with the ultrasound probe, we were not able to test at angles from 290 to 250 degrees.

Validation with different ultrasound powers, apertures, focuses, orientations, and gains
The triggering condition of the active echo system is mainly affected by two factors: the amount of acoustic power received by the active element, and the gain of the receiver. The former factor also relies on many parameters like the probe transmission power, transmission aperture, focusing, tissue attenuation, etc. An experiment is performed to investigate the working conditions of AUSPIS. In this experiment, the active echo element is aligned with the probe mid-plane in a water tank. An attenuation layer is placed between the probe and catheter to mimic tissue attenuation. The trigger count is recorded under different parameter settings. Shown in Figure 8-11, the color code indicates the trigger count per frame. Generally, to achieve the same trigger count, it requires more probe transmission power at lower receiver gain than at a higher gain. Larger transmission aperture delivers more acoustic power, thus resulting in a higher trigger count when other parameters are the same. Comparing the focusing and no focusing, the former one is relatively less sensitive to the receiver gain setting, i.e., the trigger count changes less when varying the gain. Take the figure 8 and 9 ''Tx aperture = 32'' as an example, the color bands between the contour lines are wider in 8a, which indicates that, within a larger range of receiver gain setting, the trigger count remains the same. This is because the focused beam has smaller beam width, when the ultrasound probe acquires a Bmode image frame, only the A-mode lines close to the element can trigger the active echo. Since the energy is spatially concentrated, less transmission power is required to trigger the active echo, so in the focused beam conditions, the functional zone (colored area) extends to lower gain area, as shown in figure 8. In the unfocused case, since the beams are wider and have more overlapping, the beams far away from the element may also trigger the echo, so the trigger count growths rapidly when increasing the gain. Similarly, smaller transmission aperture results in wider beam, the bands are narrower than that with the larger aperture. In the experiment, the data with a trigger count higher than 40 are discarded; that is why the colored area does not extend to the very high gain region in figure 8-11. Though it may be useful in the quick-tool-searching purpose, high gain with large trigger counts makes the active echo spot wide and distorted, which may result in an inaccurate tool indication in the original B-mode image.

Detection depth test
To further investigate the extreme functional conditions of the system, we performed two lab bench experiments. In the first one, we insert the catheter into ex vivo chicken breast tissue at 8.5 cm depth, which is close to the limit of the probe we use. This is to mimic the deep needle insertion in clinical operations. An ultrasound video is given in Video S1. In the second experiment, we placed a 1 inch thick aluminum plate in the water tank between the imaging probe and catheter to mimic the impedance of highly mismatched tissues like hard bone. As shown in figure 12, in both of the two experiments, the catheter can be effectively localized by the active echo spot. In the second test, since the speed of sound in the aluminum layer is much higher than that in water, the spot looks much closer than it actually is in the B-mode image.

Validation with different imaging modes
Most clinical ultrasound systems provide different imaging modes, like harmonic mode (or pulse inversion mode), high resolution mode, high penetration mode, etc., which also affect active echo system performance. Using the Sonix Touch system with a L14-5W linear probe, we tested the active echo under four modes: harmonic mode, general mode, resolution mode, and penetration mode. Videos of the ultrasound system display are given in Videos S2, S3, S4, and S5. In Harmonic mode, the ultrasound system transmits pulses with central frequency of 5 MHz, and reconstructs the B-mode image from the received second harmonic (10 MHz) signals. In both general and resolution mode, the transmission frequency is 10 MHz; the difference is that in the resolution mode, the Rx frequency band is slightly higher than the general mode for better image resolution. In the penetration mode, the Tx frequency is reduced to 6.6 MHz to improve the penetration depth. We compared the signal to noise ratio (SNR) and contrast to noise ratio (CNR) of chicken breast tissue images using these imaging methods. The SNR is defined as: where the ''DiffSignal'' is differential signal extracted by using the echo on frame subtracting the echo off frame, the ''background'' is the echo off frame, the ''SurroundingArea'' is a 50 by 50 pixel area around the active echo spot.
Comparing the B-mode images in figure 13, the harmonicmode image has the most observable echo spot. This is expected since the harmonic imaging relies on the tissue harmonic echoes, the amplitude of which is much lower than that of the original pulses. The active echo element is configured to transmit at a frequency close to the second harmonic (10 MHz) in this experiment, so the active echo signal significantly stands out from the tissue harmonic background. The penetration-mode image has the least observable echo spot, in part because the active echo frequency (about 9 MHz) is out of the Rx frequency band in this mode. This observation is consistent with the SNR and CNR measurements, as shown in table 1. However, even in the penetration mode, though the active echo spot is not clear in the static images, it is easy for the human eye to capture the spot during the operation due to its blinking.

Validation with different AE transmission powers
Despite the imaging methods, the driving voltage of the element also affects the active echo spot visualization, since higher driving voltage results in stronger ultrasound. As shown in figure 14, with 20 V driving pulse, the echo spot is dimmer than that with 58 V driving pulse. However, since the B-mode image does not linearly represent the signal strength, we could not see the linear relationship in the CNR and SNR measurements, as shown in table 2. Figure 15 shows an example of arbitrary pattern injection method. The left image is the reference, the B-mode image of the catheter in a water tank, and right is the image with the pattern injection system turned on. A virtual ''JHU'' pattern is shown in the B-mode image. A video is given in Video S6. In this image, the ''pixel'' that forms the pattern can be seen. Each pixel corresponds to one ultrasound pulse firing from the catheter. Method 1 mentioned in the pattern injection paragraph is used in this experiment. Ultrasound system synchronization trigger is utilized, and the pattern is injected into an absolute position in the image coordinate.

Mid-plane indication using pattern injection
In the previous paragraph, we introduced the method to detect the image mid-plane. The detection result can be directly displayed on the B-mode image using arbitrary pattern injection technique. As shown in figure 16, virtual bars that are proportional to the trigger count are injected into the image. With this real-time feedback, the operator can easily navigate the tool tip to the ultrasound image mid-plane, which is indicated by the maximum number of bars. Videos of this experiment are given in Videos S7, S8. This experiment is an implementation of method 2 mentioned in the pattern injection paragraph. Only the received signal is used for synchronization, and the pattern is injected into a relevant position moving with the active element in the image.

In vivo experiment
The in vivo experiment was performed on a pig. The catheter is inserted into the liver tissue. 3D Volumetric B-mode images are collected by a 3D wobbler probe 4DL14-5/38. Figure 17 shows the images of the liver tissue with the catheter inserted. In a) and b), the catheter incidents the B-mode image plane with a small angle. c) and d) has the catheter perpendicular with the image plane. Although the catheter is visible in these B-mode images due to its large diameter and small insertion angle, it is not very distinguishable from the tissue texture, as shown in figure 17 a). Also, it is difficult to identify the location of the element, especially in the condition of c). In figure 17 b) and d), the active echo is enabled. With the echo spot clearly shown in the image, it is easy to identify the catheter from the tissue, and locate the element position.

Discussion
The previous discussions are based on linear imaging arrays. For phased arrays, theoretically the active reflection, blinking and frequency modulation functions should have similar performance. The arbitrary pattern injection, however, works differently in the case of phased arrays or curved linear arrays due to the different image forming methods and parameters. How the system works with different type of probes is a future work to be investigated.
With single AE element, the AUSPIS provides 5 degrees of freedom to the probe, as shown in figure 18. It limits the image plane to intersect the AE element, which is a point. As a comparison, active beam-steering method constrains the image plane to intersect the catheter line, thus resulting in 4 degrees of freedom. In clinic, more DOFs mean less constraints on the probe position and orientation; this is especially important in applications like cardiac catheterization, in which doctors have to image from the very limited imaging window to avoid the ribs. Like any other tracking system, AUSPIS has its limitations. This method is not able to localize the tool or project the insertion path when the AE element is far away from the ultrasound image plane. Using catheters with single AE element, the maximum detectable range in our experiments is around 612 mm from the mid-plane. However, for certain applications, it can be improved by integrating multiple AE elements at different locations of the tool. Moreover, using multiple active elements not only the tool position but also the pose can be tracked. Another limitation in the prototype system is the shape and size of the AE element. For easy assembling and wiring, the piezoelectric element we use is a 262 mm cylindrical tube. The localization accuracy is expected to improve by reducing the element size. Last but not least, the operation of AUSPIS is very different from existing tracking methods. Although we have demonstrated the concept in both ex  vivo and in vivo experiments, further study is needed to investigate how the technology can be integrated into the clinic workflow.

Conclusions
In summary, we introduced a new active ultrasound pattern injection method for interventional tool tracking, designed and built a prototype system, and performed preliminary ex vivo and in vivo experiment to prove feasibility. Significant visualization enhancement and highly accurate mid-plane localization have been achieved. Arbitrary pattern injection has been demonstrated. We also presented the study and analysis of the system operating under different configurations. Finally the limitations and future developments have been discussed.

Supporting Information
Video S1 The active echo spot under 8.5 cm deep chicken breast tissue.